Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
A couple more points to make re: Uwe's comments: > An important point that we should keep in (and why I was a bit concerned in > the previous times this discussion was raised) is that we have to be careful > to not pull everything that touches Arrow into the Arrow repository. An important distinction here is community and development process, why I focused on the "Community over Code" rationale in my original e-mail. I think we should make decisions that optimize for the community's health and productivity over satisfying arbitrary constraints. Our community's health should be measured by our ability to author and merge changes as painlessly as possible and to be able to consistently deliver high quality software releases. IMHO we have fallen short of both of these goals > Having separate repositories for projects with each its own release cycle is > for me still the aim for the longterm. A monorepo does not imply that components cannot have separate release cycles. As an example, in the JS community a tool Lerna has developed precisely because of this https://lernajs.io/ "Splitting up large codebases into separate independently versioned packages is extremely useful for code sharing. However, making changes across many repositories is messy and difficult to track, and testing across repositories gets complicated really fast. To solve these (and many other) problems, some projects will organize their codebases into multi-package repositories. Projects like Babel, React, Angular, Ember, Meteor, Jest, and many others develop all of their packages within a single repository." I think that we have arrived exactly to the same point as these other projects. We may need to develop some of our own tooling to assist with managing our monorepo (where we are already shipping multiple components) - Wes On Sun, Aug 19, 2018 at 1:30 PM, Wes McKinney wrote: > hi Uwe, > > I agree with your points. Currently we have 3 software artifacts: > > 1. Arrow C++ libraries > 2. Parquet C++ libraries with Arrow columnar integration > 3. C++ interop layer for Python + Cython bindings > > Changes in #1 prompt an awkward workflow involving multiple PRs; as a > result of this we just recently jumped 8 months from the pinned > version of Arrow in parquet-cpp. This obviously is an antipattern. If > we had a much larger group of core developers, this might be more > maintainable > > Of course changes in #2 also impact #3; a lot of our bug reports and > feature requests are coming inbound because of #3, and we have > struggled to be able to respond to the needs of users (and other > developers like Robert Gruener who are trying to use this software in > a large data warehouse) > > There is also the release coordination issue where having users > simultaneously using a released version of both projects hasn't really > happened, so we effectively already have been treating Parquet like a > vendored component in our packaging process. > > Realistically I think once #2 has become a more functionally complete > and as a result a more slowly moving piece of software, we can > contemplate splitting out all or parts of its development process back > into another repository. I think we have a ton of work to do yet on > Parquet core, particularly optimizing for high latency storage (HDFS, > S3, GCP, etc.), and it wouldn't really make sense to do such platform > level work anywhere but #1 > > - Wes > > On Sun, Aug 19, 2018 at 8:37 AM, Uwe L. Korn wrote: >> Back from vacation, I also want to finally raise my voice. >> >> With the current state of the Parquet<->Arrow development, I see a benefit >> in merging the code base for now, but not necessarily forever. >> >> Parquet C++ is the main code base of an artefact for which an Arrow C++ >> adapter is built and that uses some of the more standard-library features of >> Arrow. It is the go-to place where also the same toolchain and CI setup is >> used. Here we also directly apply all improvements that we make in Arrow >> itself. These are the points that make it special in comparison to other >> tools providing Arrow adapters like Turbodbc. >> >> Thus, I think that the current move to merge the code bases is ok for me. I >> must say that I'm not 100% certain that this is the best move but currently >> I lack better alternatives. As previously mentioned, we should take extra >> care that we can still do separate releases and also provide a path for a >> future where we split parquet-cpp into its own project/repository again. >> >> An important point that we should keep in (and why I was a bit concerned in >> the previous times this discussion was raised) is that we have to be careful >> to not pull everything that touches Arrow into the Arrow repository. Having >> separate repositories for projects with each its own release cycle is for me >> still the aim for the longterm. I expect that there will be many more >> projects that will use Arrow's I/O libraries as well as will
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi Uwe, I agree with your points. Currently we have 3 software artifacts: 1. Arrow C++ libraries 2. Parquet C++ libraries with Arrow columnar integration 3. C++ interop layer for Python + Cython bindings Changes in #1 prompt an awkward workflow involving multiple PRs; as a result of this we just recently jumped 8 months from the pinned version of Arrow in parquet-cpp. This obviously is an antipattern. If we had a much larger group of core developers, this might be more maintainable Of course changes in #2 also impact #3; a lot of our bug reports and feature requests are coming inbound because of #3, and we have struggled to be able to respond to the needs of users (and other developers like Robert Gruener who are trying to use this software in a large data warehouse) There is also the release coordination issue where having users simultaneously using a released version of both projects hasn't really happened, so we effectively already have been treating Parquet like a vendored component in our packaging process. Realistically I think once #2 has become a more functionally complete and as a result a more slowly moving piece of software, we can contemplate splitting out all or parts of its development process back into another repository. I think we have a ton of work to do yet on Parquet core, particularly optimizing for high latency storage (HDFS, S3, GCP, etc.), and it wouldn't really make sense to do such platform level work anywhere but #1 - Wes On Sun, Aug 19, 2018 at 8:37 AM, Uwe L. Korn wrote: > Back from vacation, I also want to finally raise my voice. > > With the current state of the Parquet<->Arrow development, I see a benefit in > merging the code base for now, but not necessarily forever. > > Parquet C++ is the main code base of an artefact for which an Arrow C++ > adapter is built and that uses some of the more standard-library features of > Arrow. It is the go-to place where also the same toolchain and CI setup is > used. Here we also directly apply all improvements that we make in Arrow > itself. These are the points that make it special in comparison to other > tools providing Arrow adapters like Turbodbc. > > Thus, I think that the current move to merge the code bases is ok for me. I > must say that I'm not 100% certain that this is the best move but currently I > lack better alternatives. As previously mentioned, we should take extra care > that we can still do separate releases and also provide a path for a future > where we split parquet-cpp into its own project/repository again. > > An important point that we should keep in (and why I was a bit concerned in > the previous times this discussion was raised) is that we have to be careful > to not pull everything that touches Arrow into the Arrow repository. Having > separate repositories for projects with each its own release cycle is for me > still the aim for the longterm. I expect that there will be many more > projects that will use Arrow's I/O libraries as well as will omit Arrow > structures. These libraries should be also usable in Python/C++/Ruby/R/… > These libraries are then hopefully not all developed by the same core group > of Arrow/Parquet developers we have currently. For this to function really > well, we will need a more stable API in Arrow as well as a good set of build > tooling that other libraries can build up when using Arrow functionality. In > addition to being stable, the API must also provide a good UX in the > abstraction layers the Arrow functions are provided so that high-performance > applications are not high-maintenance due to frequent API changes in Arrow. > That said, this is currently is wish for the future. We are currently > building and iterating heavily on these APIs to form a good basis for future > developments. Thus the repo merge will hopefully improve the development > speed so that we have to spent less time on toolchain maintenance and can > focus on the user-facing APIs. > > Uwe > > On Tue, Aug 7, 2018, at 10:45 PM, Wes McKinney wrote: >> Thanks Ryan, will do. The people I'd still like to hear from are: >> >> * Phillip Cloud >> * Uwe Korn >> >> As ASF contributors we are responsible to both be pragmatic as well as >> act in the best interests of the community's health and productivity. >> >> >> >> On Tue, Aug 7, 2018 at 12:12 PM, Ryan Blue wrote: >> > I don't have an opinion here, but could someone send a summary of what is >> > decided to the dev list once there is consensus? This is a long thread for >> > parts of the project I don't work on, so I haven't followed it very >> > closely. >> > >> > On Tue, Aug 7, 2018 at 8:22 AM Wes McKinney wrote: >> > >> >> > It will be difficult to track parquet-cpp changes if they get mixed with >> >> Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? >> >> Can we enforce that parquet-cpp changes will not be committed without a >> >> corresponding Parquet JIRA? >> >> >> >> I think we would
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Back from vacation, I also want to finally raise my voice. With the current state of the Parquet<->Arrow development, I see a benefit in merging the code base for now, but not necessarily forever. Parquet C++ is the main code base of an artefact for which an Arrow C++ adapter is built and that uses some of the more standard-library features of Arrow. It is the go-to place where also the same toolchain and CI setup is used. Here we also directly apply all improvements that we make in Arrow itself. These are the points that make it special in comparison to other tools providing Arrow adapters like Turbodbc. Thus, I think that the current move to merge the code bases is ok for me. I must say that I'm not 100% certain that this is the best move but currently I lack better alternatives. As previously mentioned, we should take extra care that we can still do separate releases and also provide a path for a future where we split parquet-cpp into its own project/repository again. An important point that we should keep in (and why I was a bit concerned in the previous times this discussion was raised) is that we have to be careful to not pull everything that touches Arrow into the Arrow repository. Having separate repositories for projects with each its own release cycle is for me still the aim for the longterm. I expect that there will be many more projects that will use Arrow's I/O libraries as well as will omit Arrow structures. These libraries should be also usable in Python/C++/Ruby/R/… These libraries are then hopefully not all developed by the same core group of Arrow/Parquet developers we have currently. For this to function really well, we will need a more stable API in Arrow as well as a good set of build tooling that other libraries can build up when using Arrow functionality. In addition to being stable, the API must also provide a good UX in the abstraction layers the Arrow functions are provided so that high-performance applications are not high-maintenance due to frequent API changes in Arrow. That said, this is currently is wish for the future. We are currently building and iterating heavily on these APIs to form a good basis for future developments. Thus the repo merge will hopefully improve the development speed so that we have to spent less time on toolchain maintenance and can focus on the user-facing APIs. Uwe On Tue, Aug 7, 2018, at 10:45 PM, Wes McKinney wrote: > Thanks Ryan, will do. The people I'd still like to hear from are: > > * Phillip Cloud > * Uwe Korn > > As ASF contributors we are responsible to both be pragmatic as well as > act in the best interests of the community's health and productivity. > > > > On Tue, Aug 7, 2018 at 12:12 PM, Ryan Blue wrote: > > I don't have an opinion here, but could someone send a summary of what is > > decided to the dev list once there is consensus? This is a long thread for > > parts of the project I don't work on, so I haven't followed it very closely. > > > > On Tue, Aug 7, 2018 at 8:22 AM Wes McKinney wrote: > > > >> > It will be difficult to track parquet-cpp changes if they get mixed with > >> Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? > >> Can we enforce that parquet-cpp changes will not be committed without a > >> corresponding Parquet JIRA? > >> > >> I think we would use the following policy: > >> > >> * use PARQUET-XXX for issues relating to Parquet core > >> * use ARROW-XXX for issues relation to Arrow's consumption of Parquet > >> core (e.g. changes that are in parquet/arrow right now) > >> > >> We've already been dealing with annoyances relating to issues > >> straddling the two projects (debugging an issue on Arrow side to find > >> that it has to be fixed on Parquet side); this would make things > >> simpler for us > >> > >> > I would also like to keep changes to parquet-cpp on a separate commit to > >> simplify forking later (if needed) and be able to maintain the commit > >> history. I don't know if its possible to squash parquet-cpp commits and > >> arrow commits separately before merging. > >> > >> This seems rather onerous for both contributors and maintainers and > >> not in line with the goal of improving productivity. In the event that > >> we fork I see it as a traumatic event for the community. If it does > >> happen, then we can write a script (using git filter-branch and other > >> such tools) to extract commits related to the forked code. > >> > >> - Wes > >> > >> On Tue, Aug 7, 2018 at 10:37 AM, Deepak Majeti > >> wrote: > >> > I have a few more logistical questions to add. > >> > > >> > It will be difficult to track parquet-cpp changes if they get mixed with > >> > Arrow changes. Will we establish some guidelines for filing Parquet > >> JIRAs? > >> > Can we enforce that parquet-cpp changes will not be committed without a > >> > corresponding Parquet JIRA? > >> > > >> > I would also like to keep changes to parquet-cpp on a separate commit to > >> >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Thanks Ryan, will do. The people I'd still like to hear from are: * Phillip Cloud * Uwe Korn As ASF contributors we are responsible to both be pragmatic as well as act in the best interests of the community's health and productivity. On Tue, Aug 7, 2018 at 12:12 PM, Ryan Blue wrote: > I don't have an opinion here, but could someone send a summary of what is > decided to the dev list once there is consensus? This is a long thread for > parts of the project I don't work on, so I haven't followed it very closely. > > On Tue, Aug 7, 2018 at 8:22 AM Wes McKinney wrote: > >> > It will be difficult to track parquet-cpp changes if they get mixed with >> Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? >> Can we enforce that parquet-cpp changes will not be committed without a >> corresponding Parquet JIRA? >> >> I think we would use the following policy: >> >> * use PARQUET-XXX for issues relating to Parquet core >> * use ARROW-XXX for issues relation to Arrow's consumption of Parquet >> core (e.g. changes that are in parquet/arrow right now) >> >> We've already been dealing with annoyances relating to issues >> straddling the two projects (debugging an issue on Arrow side to find >> that it has to be fixed on Parquet side); this would make things >> simpler for us >> >> > I would also like to keep changes to parquet-cpp on a separate commit to >> simplify forking later (if needed) and be able to maintain the commit >> history. I don't know if its possible to squash parquet-cpp commits and >> arrow commits separately before merging. >> >> This seems rather onerous for both contributors and maintainers and >> not in line with the goal of improving productivity. In the event that >> we fork I see it as a traumatic event for the community. If it does >> happen, then we can write a script (using git filter-branch and other >> such tools) to extract commits related to the forked code. >> >> - Wes >> >> On Tue, Aug 7, 2018 at 10:37 AM, Deepak Majeti >> wrote: >> > I have a few more logistical questions to add. >> > >> > It will be difficult to track parquet-cpp changes if they get mixed with >> > Arrow changes. Will we establish some guidelines for filing Parquet >> JIRAs? >> > Can we enforce that parquet-cpp changes will not be committed without a >> > corresponding Parquet JIRA? >> > >> > I would also like to keep changes to parquet-cpp on a separate commit to >> > simplify forking later (if needed) and be able to maintain the commit >> > history. I don't know if its possible to squash parquet-cpp commits and >> > arrow commits separately before merging. >> > >> > >> > On Tue, Aug 7, 2018 at 8:57 AM Wes McKinney wrote: >> > >> >> Do other people have opinions? I would like to undertake this work in >> >> the near future (the next 8-10 weeks); I would be OK with taking >> >> responsibility for the primary codebase surgery. >> >> >> >> Some logistical questions: >> >> >> >> * We have a handful of pull requests in flight in parquet-cpp that >> >> would need to be resolved / merged >> >> * We should probably cut a status-quo cpp-1.5.0 release, with future >> >> releases cut out of the new structure >> >> * Management of shared commit rights (I can discuss with the Arrow >> >> PMC; I believe that approving any committer who has actively >> >> maintained parquet-cpp should be a reasonable approach per Ted's >> >> comments) >> >> >> >> If working more closely together proves to not be working out after >> >> some period of time, I will be fully supportive of a fork or something >> >> like it >> >> >> >> Thanks, >> >> Wes >> >> >> >> On Wed, Aug 1, 2018 at 3:39 PM, Wes McKinney >> wrote: >> >> > Thanks Tim. >> >> > >> >> > Indeed, it's not very simple. Just today Antoine cleaned up some >> >> > platform code intending to improve the performance of bit-packing in >> >> > Parquet writes, and we resulted with 2 interdependent PRs >> >> > >> >> > * https://github.com/apache/parquet-cpp/pull/483 >> >> > * https://github.com/apache/arrow/pull/2355 >> >> > >> >> > Changes that impact the Python interface to Parquet are even more >> >> complex. >> >> > >> >> > Adding options to Arrow's CMake build system to only build >> >> > Parquet-related code and dependencies (in a monorepo framework) would >> >> > not be difficult, and amount to writing "make parquet". >> >> > >> >> > See e.g. https://stackoverflow.com/a/17201375. The desired commands >> to >> >> > build and install the Parquet core libraries and their dependencies >> >> > would be: >> >> > >> >> > ninja parquet && ninja install >> >> > >> >> > - Wes >> >> > >> >> > On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong >> >> > wrote: >> >> >> I don't have a direct stake in this beyond wanting to see Parquet be >> >> >> successful, but I thought I'd give my two cents. >> >> >> >> >> >> For me, the thing that makes the biggest difference in contributing >> to a >> >> >> new codebase is the number of steps in the workflow for writing, >> >> testing,
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I don't have an opinion here, but could someone send a summary of what is decided to the dev list once there is consensus? This is a long thread for parts of the project I don't work on, so I haven't followed it very closely. On Tue, Aug 7, 2018 at 8:22 AM Wes McKinney wrote: > > It will be difficult to track parquet-cpp changes if they get mixed with > Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? > Can we enforce that parquet-cpp changes will not be committed without a > corresponding Parquet JIRA? > > I think we would use the following policy: > > * use PARQUET-XXX for issues relating to Parquet core > * use ARROW-XXX for issues relation to Arrow's consumption of Parquet > core (e.g. changes that are in parquet/arrow right now) > > We've already been dealing with annoyances relating to issues > straddling the two projects (debugging an issue on Arrow side to find > that it has to be fixed on Parquet side); this would make things > simpler for us > > > I would also like to keep changes to parquet-cpp on a separate commit to > simplify forking later (if needed) and be able to maintain the commit > history. I don't know if its possible to squash parquet-cpp commits and > arrow commits separately before merging. > > This seems rather onerous for both contributors and maintainers and > not in line with the goal of improving productivity. In the event that > we fork I see it as a traumatic event for the community. If it does > happen, then we can write a script (using git filter-branch and other > such tools) to extract commits related to the forked code. > > - Wes > > On Tue, Aug 7, 2018 at 10:37 AM, Deepak Majeti > wrote: > > I have a few more logistical questions to add. > > > > It will be difficult to track parquet-cpp changes if they get mixed with > > Arrow changes. Will we establish some guidelines for filing Parquet > JIRAs? > > Can we enforce that parquet-cpp changes will not be committed without a > > corresponding Parquet JIRA? > > > > I would also like to keep changes to parquet-cpp on a separate commit to > > simplify forking later (if needed) and be able to maintain the commit > > history. I don't know if its possible to squash parquet-cpp commits and > > arrow commits separately before merging. > > > > > > On Tue, Aug 7, 2018 at 8:57 AM Wes McKinney wrote: > > > >> Do other people have opinions? I would like to undertake this work in > >> the near future (the next 8-10 weeks); I would be OK with taking > >> responsibility for the primary codebase surgery. > >> > >> Some logistical questions: > >> > >> * We have a handful of pull requests in flight in parquet-cpp that > >> would need to be resolved / merged > >> * We should probably cut a status-quo cpp-1.5.0 release, with future > >> releases cut out of the new structure > >> * Management of shared commit rights (I can discuss with the Arrow > >> PMC; I believe that approving any committer who has actively > >> maintained parquet-cpp should be a reasonable approach per Ted's > >> comments) > >> > >> If working more closely together proves to not be working out after > >> some period of time, I will be fully supportive of a fork or something > >> like it > >> > >> Thanks, > >> Wes > >> > >> On Wed, Aug 1, 2018 at 3:39 PM, Wes McKinney > wrote: > >> > Thanks Tim. > >> > > >> > Indeed, it's not very simple. Just today Antoine cleaned up some > >> > platform code intending to improve the performance of bit-packing in > >> > Parquet writes, and we resulted with 2 interdependent PRs > >> > > >> > * https://github.com/apache/parquet-cpp/pull/483 > >> > * https://github.com/apache/arrow/pull/2355 > >> > > >> > Changes that impact the Python interface to Parquet are even more > >> complex. > >> > > >> > Adding options to Arrow's CMake build system to only build > >> > Parquet-related code and dependencies (in a monorepo framework) would > >> > not be difficult, and amount to writing "make parquet". > >> > > >> > See e.g. https://stackoverflow.com/a/17201375. The desired commands > to > >> > build and install the Parquet core libraries and their dependencies > >> > would be: > >> > > >> > ninja parquet && ninja install > >> > > >> > - Wes > >> > > >> > On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong > >> > wrote: > >> >> I don't have a direct stake in this beyond wanting to see Parquet be > >> >> successful, but I thought I'd give my two cents. > >> >> > >> >> For me, the thing that makes the biggest difference in contributing > to a > >> >> new codebase is the number of steps in the workflow for writing, > >> testing, > >> >> posting and iterating on a commit and also the number of > opportunities > >> for > >> >> missteps. The size of the repo and build/test times matter but are > >> >> secondary so long as the workflow is simple and reliable. > >> >> > >> >> I don't really know what the current state of things is, but it > sounds > >> like > >> >> it's not as simple as check out -> build -> test if you're doing a >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
> It will be difficult to track parquet-cpp changes if they get mixed with > Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? > Can we enforce that parquet-cpp changes will not be committed without a > corresponding Parquet JIRA? I think we would use the following policy: * use PARQUET-XXX for issues relating to Parquet core * use ARROW-XXX for issues relation to Arrow's consumption of Parquet core (e.g. changes that are in parquet/arrow right now) We've already been dealing with annoyances relating to issues straddling the two projects (debugging an issue on Arrow side to find that it has to be fixed on Parquet side); this would make things simpler for us > I would also like to keep changes to parquet-cpp on a separate commit to > simplify forking later (if needed) and be able to maintain the commit > history. I don't know if its possible to squash parquet-cpp commits and > arrow commits separately before merging. This seems rather onerous for both contributors and maintainers and not in line with the goal of improving productivity. In the event that we fork I see it as a traumatic event for the community. If it does happen, then we can write a script (using git filter-branch and other such tools) to extract commits related to the forked code. - Wes On Tue, Aug 7, 2018 at 10:37 AM, Deepak Majeti wrote: > I have a few more logistical questions to add. > > It will be difficult to track parquet-cpp changes if they get mixed with > Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? > Can we enforce that parquet-cpp changes will not be committed without a > corresponding Parquet JIRA? > > I would also like to keep changes to parquet-cpp on a separate commit to > simplify forking later (if needed) and be able to maintain the commit > history. I don't know if its possible to squash parquet-cpp commits and > arrow commits separately before merging. > > > On Tue, Aug 7, 2018 at 8:57 AM Wes McKinney wrote: > >> Do other people have opinions? I would like to undertake this work in >> the near future (the next 8-10 weeks); I would be OK with taking >> responsibility for the primary codebase surgery. >> >> Some logistical questions: >> >> * We have a handful of pull requests in flight in parquet-cpp that >> would need to be resolved / merged >> * We should probably cut a status-quo cpp-1.5.0 release, with future >> releases cut out of the new structure >> * Management of shared commit rights (I can discuss with the Arrow >> PMC; I believe that approving any committer who has actively >> maintained parquet-cpp should be a reasonable approach per Ted's >> comments) >> >> If working more closely together proves to not be working out after >> some period of time, I will be fully supportive of a fork or something >> like it >> >> Thanks, >> Wes >> >> On Wed, Aug 1, 2018 at 3:39 PM, Wes McKinney wrote: >> > Thanks Tim. >> > >> > Indeed, it's not very simple. Just today Antoine cleaned up some >> > platform code intending to improve the performance of bit-packing in >> > Parquet writes, and we resulted with 2 interdependent PRs >> > >> > * https://github.com/apache/parquet-cpp/pull/483 >> > * https://github.com/apache/arrow/pull/2355 >> > >> > Changes that impact the Python interface to Parquet are even more >> complex. >> > >> > Adding options to Arrow's CMake build system to only build >> > Parquet-related code and dependencies (in a monorepo framework) would >> > not be difficult, and amount to writing "make parquet". >> > >> > See e.g. https://stackoverflow.com/a/17201375. The desired commands to >> > build and install the Parquet core libraries and their dependencies >> > would be: >> > >> > ninja parquet && ninja install >> > >> > - Wes >> > >> > On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong >> > wrote: >> >> I don't have a direct stake in this beyond wanting to see Parquet be >> >> successful, but I thought I'd give my two cents. >> >> >> >> For me, the thing that makes the biggest difference in contributing to a >> >> new codebase is the number of steps in the workflow for writing, >> testing, >> >> posting and iterating on a commit and also the number of opportunities >> for >> >> missteps. The size of the repo and build/test times matter but are >> >> secondary so long as the workflow is simple and reliable. >> >> >> >> I don't really know what the current state of things is, but it sounds >> like >> >> it's not as simple as check out -> build -> test if you're doing a >> >> cross-repo change. Circular dependencies are a real headache. >> >> >> >> On Tue, Jul 31, 2018 at 2:44 PM, Wes McKinney >> wrote: >> >> >> >>> hi, >> >>> >> >>> On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti < >> majeti.dee...@gmail.com> >> >>> wrote: >> >>> > I think the circular dependency can be broken if we build a new >> library >> >>> for >> >>> > the platform code. This will also make it easy for other projects >> such as >> >>> > ORC to use it. >> >>> > I also
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I have a few more logistical questions to add. It will be difficult to track parquet-cpp changes if they get mixed with Arrow changes. Will we establish some guidelines for filing Parquet JIRAs? Can we enforce that parquet-cpp changes will not be committed without a corresponding Parquet JIRA? I would also like to keep changes to parquet-cpp on a separate commit to simplify forking later (if needed) and be able to maintain the commit history. I don't know if its possible to squash parquet-cpp commits and arrow commits separately before merging. On Tue, Aug 7, 2018 at 8:57 AM Wes McKinney wrote: > Do other people have opinions? I would like to undertake this work in > the near future (the next 8-10 weeks); I would be OK with taking > responsibility for the primary codebase surgery. > > Some logistical questions: > > * We have a handful of pull requests in flight in parquet-cpp that > would need to be resolved / merged > * We should probably cut a status-quo cpp-1.5.0 release, with future > releases cut out of the new structure > * Management of shared commit rights (I can discuss with the Arrow > PMC; I believe that approving any committer who has actively > maintained parquet-cpp should be a reasonable approach per Ted's > comments) > > If working more closely together proves to not be working out after > some period of time, I will be fully supportive of a fork or something > like it > > Thanks, > Wes > > On Wed, Aug 1, 2018 at 3:39 PM, Wes McKinney wrote: > > Thanks Tim. > > > > Indeed, it's not very simple. Just today Antoine cleaned up some > > platform code intending to improve the performance of bit-packing in > > Parquet writes, and we resulted with 2 interdependent PRs > > > > * https://github.com/apache/parquet-cpp/pull/483 > > * https://github.com/apache/arrow/pull/2355 > > > > Changes that impact the Python interface to Parquet are even more > complex. > > > > Adding options to Arrow's CMake build system to only build > > Parquet-related code and dependencies (in a monorepo framework) would > > not be difficult, and amount to writing "make parquet". > > > > See e.g. https://stackoverflow.com/a/17201375. The desired commands to > > build and install the Parquet core libraries and their dependencies > > would be: > > > > ninja parquet && ninja install > > > > - Wes > > > > On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong > > wrote: > >> I don't have a direct stake in this beyond wanting to see Parquet be > >> successful, but I thought I'd give my two cents. > >> > >> For me, the thing that makes the biggest difference in contributing to a > >> new codebase is the number of steps in the workflow for writing, > testing, > >> posting and iterating on a commit and also the number of opportunities > for > >> missteps. The size of the repo and build/test times matter but are > >> secondary so long as the workflow is simple and reliable. > >> > >> I don't really know what the current state of things is, but it sounds > like > >> it's not as simple as check out -> build -> test if you're doing a > >> cross-repo change. Circular dependencies are a real headache. > >> > >> On Tue, Jul 31, 2018 at 2:44 PM, Wes McKinney > wrote: > >> > >>> hi, > >>> > >>> On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti < > majeti.dee...@gmail.com> > >>> wrote: > >>> > I think the circular dependency can be broken if we build a new > library > >>> for > >>> > the platform code. This will also make it easy for other projects > such as > >>> > ORC to use it. > >>> > I also remember your proposal a while ago of having a separate > project > >>> for > >>> > the platform code. That project can live in the arrow repo. > However, one > >>> > has to clone the entire apache arrow repo but can just build the > platform > >>> > code. This will be temporary until we can find a new home for it. > >>> > > >>> > The dependency will look like: > >>> > libarrow(arrow core / bindings) <- libparquet (parquet core) <- > >>> > libplatform(platform api) > >>> > > >>> > CI workflow will clone the arrow project twice, once for the platform > >>> > library and once for the arrow-core/bindings library. > >>> > >>> This seems like an interesting proposal; the best place to work toward > >>> this goal (if it is even possible; the build system interactions and > >>> ASF release management are the hard problems) is to have all of the > >>> code in a single repository. ORC could already be using Arrow if it > >>> wanted, but the ORC contributors aren't active in Arrow. > >>> > >>> > > >>> > There is no doubt that the collaborations between the Arrow and > Parquet > >>> > communities so far have been very successful. > >>> > The reason to maintain this relationship moving forward is to > continue to > >>> > reap the mutual benefits. > >>> > We should continue to take advantage of sharing code as well. > However, I > >>> > don't see any code sharing opportunities between arrow-core and the > >>> > parquet-core. Both have different functions. > >>> >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Do other people have opinions? I would like to undertake this work in the near future (the next 8-10 weeks); I would be OK with taking responsibility for the primary codebase surgery. Some logistical questions: * We have a handful of pull requests in flight in parquet-cpp that would need to be resolved / merged * We should probably cut a status-quo cpp-1.5.0 release, with future releases cut out of the new structure * Management of shared commit rights (I can discuss with the Arrow PMC; I believe that approving any committer who has actively maintained parquet-cpp should be a reasonable approach per Ted's comments) If working more closely together proves to not be working out after some period of time, I will be fully supportive of a fork or something like it Thanks, Wes On Wed, Aug 1, 2018 at 3:39 PM, Wes McKinney wrote: > Thanks Tim. > > Indeed, it's not very simple. Just today Antoine cleaned up some > platform code intending to improve the performance of bit-packing in > Parquet writes, and we resulted with 2 interdependent PRs > > * https://github.com/apache/parquet-cpp/pull/483 > * https://github.com/apache/arrow/pull/2355 > > Changes that impact the Python interface to Parquet are even more complex. > > Adding options to Arrow's CMake build system to only build > Parquet-related code and dependencies (in a monorepo framework) would > not be difficult, and amount to writing "make parquet". > > See e.g. https://stackoverflow.com/a/17201375. The desired commands to > build and install the Parquet core libraries and their dependencies > would be: > > ninja parquet && ninja install > > - Wes > > On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong > wrote: >> I don't have a direct stake in this beyond wanting to see Parquet be >> successful, but I thought I'd give my two cents. >> >> For me, the thing that makes the biggest difference in contributing to a >> new codebase is the number of steps in the workflow for writing, testing, >> posting and iterating on a commit and also the number of opportunities for >> missteps. The size of the repo and build/test times matter but are >> secondary so long as the workflow is simple and reliable. >> >> I don't really know what the current state of things is, but it sounds like >> it's not as simple as check out -> build -> test if you're doing a >> cross-repo change. Circular dependencies are a real headache. >> >> On Tue, Jul 31, 2018 at 2:44 PM, Wes McKinney wrote: >> >>> hi, >>> >>> On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti >>> wrote: >>> > I think the circular dependency can be broken if we build a new library >>> for >>> > the platform code. This will also make it easy for other projects such as >>> > ORC to use it. >>> > I also remember your proposal a while ago of having a separate project >>> for >>> > the platform code. That project can live in the arrow repo. However, one >>> > has to clone the entire apache arrow repo but can just build the platform >>> > code. This will be temporary until we can find a new home for it. >>> > >>> > The dependency will look like: >>> > libarrow(arrow core / bindings) <- libparquet (parquet core) <- >>> > libplatform(platform api) >>> > >>> > CI workflow will clone the arrow project twice, once for the platform >>> > library and once for the arrow-core/bindings library. >>> >>> This seems like an interesting proposal; the best place to work toward >>> this goal (if it is even possible; the build system interactions and >>> ASF release management are the hard problems) is to have all of the >>> code in a single repository. ORC could already be using Arrow if it >>> wanted, but the ORC contributors aren't active in Arrow. >>> >>> > >>> > There is no doubt that the collaborations between the Arrow and Parquet >>> > communities so far have been very successful. >>> > The reason to maintain this relationship moving forward is to continue to >>> > reap the mutual benefits. >>> > We should continue to take advantage of sharing code as well. However, I >>> > don't see any code sharing opportunities between arrow-core and the >>> > parquet-core. Both have different functions. >>> >>> I think you mean the Arrow columnar format. The Arrow columnar format >>> is only one part of a project that has become quite large already >>> (https://www.slideshare.net/wesm/apache-arrow-crosslanguage-development- >>> platform-for-inmemory-data-105427919). >>> >>> > >>> > We are at a point where the parquet-cpp public API is pretty stable. We >>> > already passed that difficult stage. My take at arrow and parquet is to >>> > keep them nimble since we can. >>> >>> I believe that parquet-core has progress to make yet ahead of it. We >>> have done little work in asynchronous IO and concurrency which would >>> yield both improved read and write throughput. This aligns well with >>> other concurrency and async-IO work planned in the Arrow platform. I >>> believe that more development will happen on parquet-core once the >>> development process
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Thanks Tim. Indeed, it's not very simple. Just today Antoine cleaned up some platform code intending to improve the performance of bit-packing in Parquet writes, and we resulted with 2 interdependent PRs * https://github.com/apache/parquet-cpp/pull/483 * https://github.com/apache/arrow/pull/2355 Changes that impact the Python interface to Parquet are even more complex. Adding options to Arrow's CMake build system to only build Parquet-related code and dependencies (in a monorepo framework) would not be difficult, and amount to writing "make parquet". See e.g. https://stackoverflow.com/a/17201375. The desired commands to build and install the Parquet core libraries and their dependencies would be: ninja parquet && ninja install - Wes On Wed, Aug 1, 2018 at 2:34 PM, Tim Armstrong wrote: > I don't have a direct stake in this beyond wanting to see Parquet be > successful, but I thought I'd give my two cents. > > For me, the thing that makes the biggest difference in contributing to a > new codebase is the number of steps in the workflow for writing, testing, > posting and iterating on a commit and also the number of opportunities for > missteps. The size of the repo and build/test times matter but are > secondary so long as the workflow is simple and reliable. > > I don't really know what the current state of things is, but it sounds like > it's not as simple as check out -> build -> test if you're doing a > cross-repo change. Circular dependencies are a real headache. > > On Tue, Jul 31, 2018 at 2:44 PM, Wes McKinney wrote: > >> hi, >> >> On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti >> wrote: >> > I think the circular dependency can be broken if we build a new library >> for >> > the platform code. This will also make it easy for other projects such as >> > ORC to use it. >> > I also remember your proposal a while ago of having a separate project >> for >> > the platform code. That project can live in the arrow repo. However, one >> > has to clone the entire apache arrow repo but can just build the platform >> > code. This will be temporary until we can find a new home for it. >> > >> > The dependency will look like: >> > libarrow(arrow core / bindings) <- libparquet (parquet core) <- >> > libplatform(platform api) >> > >> > CI workflow will clone the arrow project twice, once for the platform >> > library and once for the arrow-core/bindings library. >> >> This seems like an interesting proposal; the best place to work toward >> this goal (if it is even possible; the build system interactions and >> ASF release management are the hard problems) is to have all of the >> code in a single repository. ORC could already be using Arrow if it >> wanted, but the ORC contributors aren't active in Arrow. >> >> > >> > There is no doubt that the collaborations between the Arrow and Parquet >> > communities so far have been very successful. >> > The reason to maintain this relationship moving forward is to continue to >> > reap the mutual benefits. >> > We should continue to take advantage of sharing code as well. However, I >> > don't see any code sharing opportunities between arrow-core and the >> > parquet-core. Both have different functions. >> >> I think you mean the Arrow columnar format. The Arrow columnar format >> is only one part of a project that has become quite large already >> (https://www.slideshare.net/wesm/apache-arrow-crosslanguage-development- >> platform-for-inmemory-data-105427919). >> >> > >> > We are at a point where the parquet-cpp public API is pretty stable. We >> > already passed that difficult stage. My take at arrow and parquet is to >> > keep them nimble since we can. >> >> I believe that parquet-core has progress to make yet ahead of it. We >> have done little work in asynchronous IO and concurrency which would >> yield both improved read and write throughput. This aligns well with >> other concurrency and async-IO work planned in the Arrow platform. I >> believe that more development will happen on parquet-core once the >> development process issues are resolved by having a single codebase, >> single build system, and a single CI framework. >> >> I have some gripes about design decisions made early in parquet-cpp, >> like the use of C++ exceptions. So while "stability" is a reasonable >> goal I think we should still be open to making significant changes in >> the interest of long term progress. >> >> Having now worked on these projects for more than 2 and a half years >> and the most frequent contributor to both codebases, I'm sadly far >> past the "breaking point" and not willing to continue contributing in >> a significant way to parquet-cpp if the projects remained structured >> as they are now. It's hampering progress and not serving the >> community. >> >> - Wes >> >> > >> > >> > >> > >> > On Tue, Jul 31, 2018 at 3:17 PM Wes McKinney >> wrote: >> > >> >> > The current Arrow adaptor code for parquet should live in the arrow >> >> repo. That will remove a majority of the
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I don't have a direct stake in this beyond wanting to see Parquet be successful, but I thought I'd give my two cents. For me, the thing that makes the biggest difference in contributing to a new codebase is the number of steps in the workflow for writing, testing, posting and iterating on a commit and also the number of opportunities for missteps. The size of the repo and build/test times matter but are secondary so long as the workflow is simple and reliable. I don't really know what the current state of things is, but it sounds like it's not as simple as check out -> build -> test if you're doing a cross-repo change. Circular dependencies are a real headache. On Tue, Jul 31, 2018 at 2:44 PM, Wes McKinney wrote: > hi, > > On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti > wrote: > > I think the circular dependency can be broken if we build a new library > for > > the platform code. This will also make it easy for other projects such as > > ORC to use it. > > I also remember your proposal a while ago of having a separate project > for > > the platform code. That project can live in the arrow repo. However, one > > has to clone the entire apache arrow repo but can just build the platform > > code. This will be temporary until we can find a new home for it. > > > > The dependency will look like: > > libarrow(arrow core / bindings) <- libparquet (parquet core) <- > > libplatform(platform api) > > > > CI workflow will clone the arrow project twice, once for the platform > > library and once for the arrow-core/bindings library. > > This seems like an interesting proposal; the best place to work toward > this goal (if it is even possible; the build system interactions and > ASF release management are the hard problems) is to have all of the > code in a single repository. ORC could already be using Arrow if it > wanted, but the ORC contributors aren't active in Arrow. > > > > > There is no doubt that the collaborations between the Arrow and Parquet > > communities so far have been very successful. > > The reason to maintain this relationship moving forward is to continue to > > reap the mutual benefits. > > We should continue to take advantage of sharing code as well. However, I > > don't see any code sharing opportunities between arrow-core and the > > parquet-core. Both have different functions. > > I think you mean the Arrow columnar format. The Arrow columnar format > is only one part of a project that has become quite large already > (https://www.slideshare.net/wesm/apache-arrow-crosslanguage-development- > platform-for-inmemory-data-105427919). > > > > > We are at a point where the parquet-cpp public API is pretty stable. We > > already passed that difficult stage. My take at arrow and parquet is to > > keep them nimble since we can. > > I believe that parquet-core has progress to make yet ahead of it. We > have done little work in asynchronous IO and concurrency which would > yield both improved read and write throughput. This aligns well with > other concurrency and async-IO work planned in the Arrow platform. I > believe that more development will happen on parquet-core once the > development process issues are resolved by having a single codebase, > single build system, and a single CI framework. > > I have some gripes about design decisions made early in parquet-cpp, > like the use of C++ exceptions. So while "stability" is a reasonable > goal I think we should still be open to making significant changes in > the interest of long term progress. > > Having now worked on these projects for more than 2 and a half years > and the most frequent contributor to both codebases, I'm sadly far > past the "breaking point" and not willing to continue contributing in > a significant way to parquet-cpp if the projects remained structured > as they are now. It's hampering progress and not serving the > community. > > - Wes > > > > > > > > > > > On Tue, Jul 31, 2018 at 3:17 PM Wes McKinney > wrote: > > > >> > The current Arrow adaptor code for parquet should live in the arrow > >> repo. That will remove a majority of the dependency issues. Joshua's > work > >> would not have been blocked in parquet-cpp if that adapter was in the > arrow > >> repo. This will be similar to the ORC adaptor. > >> > >> This has been suggested before, but I don't see how it would alleviate > >> any issues because of the significant dependencies on other parts of > >> the Arrow codebase. What you are proposing is: > >> > >> - (Arrow) arrow platform > >> - (Parquet) parquet core > >> - (Arrow) arrow columnar-parquet adapter interface > >> - (Arrow) Python bindings > >> > >> To make this work, somehow Arrow core / libarrow would have to be > >> built before invoking the Parquet core part of the build system. You > >> would need to pass dependent targets across different CMake build > >> systems; I don't know if it's possible (I spent some time looking into > >> it earlier this year). This is what I meant by the lack of a "concrete > >>
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi, On Tue, Jul 31, 2018 at 4:56 PM, Deepak Majeti wrote: > I think the circular dependency can be broken if we build a new library for > the platform code. This will also make it easy for other projects such as > ORC to use it. > I also remember your proposal a while ago of having a separate project for > the platform code. That project can live in the arrow repo. However, one > has to clone the entire apache arrow repo but can just build the platform > code. This will be temporary until we can find a new home for it. > > The dependency will look like: > libarrow(arrow core / bindings) <- libparquet (parquet core) <- > libplatform(platform api) > > CI workflow will clone the arrow project twice, once for the platform > library and once for the arrow-core/bindings library. This seems like an interesting proposal; the best place to work toward this goal (if it is even possible; the build system interactions and ASF release management are the hard problems) is to have all of the code in a single repository. ORC could already be using Arrow if it wanted, but the ORC contributors aren't active in Arrow. > > There is no doubt that the collaborations between the Arrow and Parquet > communities so far have been very successful. > The reason to maintain this relationship moving forward is to continue to > reap the mutual benefits. > We should continue to take advantage of sharing code as well. However, I > don't see any code sharing opportunities between arrow-core and the > parquet-core. Both have different functions. I think you mean the Arrow columnar format. The Arrow columnar format is only one part of a project that has become quite large already (https://www.slideshare.net/wesm/apache-arrow-crosslanguage-development-platform-for-inmemory-data-105427919). > > We are at a point where the parquet-cpp public API is pretty stable. We > already passed that difficult stage. My take at arrow and parquet is to > keep them nimble since we can. I believe that parquet-core has progress to make yet ahead of it. We have done little work in asynchronous IO and concurrency which would yield both improved read and write throughput. This aligns well with other concurrency and async-IO work planned in the Arrow platform. I believe that more development will happen on parquet-core once the development process issues are resolved by having a single codebase, single build system, and a single CI framework. I have some gripes about design decisions made early in parquet-cpp, like the use of C++ exceptions. So while "stability" is a reasonable goal I think we should still be open to making significant changes in the interest of long term progress. Having now worked on these projects for more than 2 and a half years and the most frequent contributor to both codebases, I'm sadly far past the "breaking point" and not willing to continue contributing in a significant way to parquet-cpp if the projects remained structured as they are now. It's hampering progress and not serving the community. - Wes > > > > > On Tue, Jul 31, 2018 at 3:17 PM Wes McKinney wrote: > >> > The current Arrow adaptor code for parquet should live in the arrow >> repo. That will remove a majority of the dependency issues. Joshua's work >> would not have been blocked in parquet-cpp if that adapter was in the arrow >> repo. This will be similar to the ORC adaptor. >> >> This has been suggested before, but I don't see how it would alleviate >> any issues because of the significant dependencies on other parts of >> the Arrow codebase. What you are proposing is: >> >> - (Arrow) arrow platform >> - (Parquet) parquet core >> - (Arrow) arrow columnar-parquet adapter interface >> - (Arrow) Python bindings >> >> To make this work, somehow Arrow core / libarrow would have to be >> built before invoking the Parquet core part of the build system. You >> would need to pass dependent targets across different CMake build >> systems; I don't know if it's possible (I spent some time looking into >> it earlier this year). This is what I meant by the lack of a "concrete >> and actionable plan". The only thing that would really work would be >> for the Parquet core to be "included" in the Arrow build system >> somehow rather than using ExternalProject. Currently Parquet builds >> Arrow using ExternalProject, and Parquet is unknown to the Arrow build >> system because it's only depended upon by the Python bindings. >> >> And even if a solution could be devised, it would not wholly resolve >> the CI workflow issues. >> >> You could make Parquet completely independent of the Arrow codebase, >> but at that point there is little reason to maintain a relationship >> between the projects or their communities. We have spent a great deal >> of effort refactoring the two projects to enable as much code sharing >> as there is now. >> >> - Wes >> >> On Tue, Jul 31, 2018 at 2:29 PM, Wes McKinney wrote: >> >> If you still strongly feel that the only way forward is to clone the >>
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
> The current Arrow adaptor code for parquet should live in the arrow repo. > That will remove a majority of the dependency issues. Joshua's work would not > have been blocked in parquet-cpp if that adapter was in the arrow repo. This > will be similar to the ORC adaptor. This has been suggested before, but I don't see how it would alleviate any issues because of the significant dependencies on other parts of the Arrow codebase. What you are proposing is: - (Arrow) arrow platform - (Parquet) parquet core - (Arrow) arrow columnar-parquet adapter interface - (Arrow) Python bindings To make this work, somehow Arrow core / libarrow would have to be built before invoking the Parquet core part of the build system. You would need to pass dependent targets across different CMake build systems; I don't know if it's possible (I spent some time looking into it earlier this year). This is what I meant by the lack of a "concrete and actionable plan". The only thing that would really work would be for the Parquet core to be "included" in the Arrow build system somehow rather than using ExternalProject. Currently Parquet builds Arrow using ExternalProject, and Parquet is unknown to the Arrow build system because it's only depended upon by the Python bindings. And even if a solution could be devised, it would not wholly resolve the CI workflow issues. You could make Parquet completely independent of the Arrow codebase, but at that point there is little reason to maintain a relationship between the projects or their communities. We have spent a great deal of effort refactoring the two projects to enable as much code sharing as there is now. - Wes On Tue, Jul 31, 2018 at 2:29 PM, Wes McKinney wrote: >> If you still strongly feel that the only way forward is to clone the >> parquet-cpp repo and part ways, I will withdraw my concern. Having two >> parquet-cpp repos is no way a better approach. > > Yes, indeed. In my view, the next best option after a monorepo is to > fork. That would obviously be a bad outcome for the community. > > It doesn't look like I will be able to convince you that a monorepo is > a good idea; what I would ask instead is that you be willing to give > it a shot, and if it turns out in the way you're describing (which I > don't think it will) then I suggest that we fork at that point. > > - Wes > > On Tue, Jul 31, 2018 at 2:14 PM, Deepak Majeti > wrote: >> Wes, >> >> Unfortunately, I cannot show you any practical fact-based problems of a >> non-existent Arrow-Parquet mono-repo. >> Bringing in related Apache community experiences are more meaningful than >> how mono-repos work at Google and other big organizations. >> We solely depend on volunteers and cannot hire full-time developers. >> You are very well aware of how difficult it has been to find more >> contributors and maintainers for Arrow. parquet-cpp already has a low >> contribution rate to its core components. >> >> We should target to ensure that new volunteers who want to contribute >> bug-fixes/features should spend the least amount of time in figuring out >> the project repo. We can never come up with an automated build system that >> caters to every possible environment. >> My only concern is if the mono-repo will make it harder for new developers >> to work on parquet-cpp core just due to the additional code, build and test >> dependencies. >> I am not saying that the Arrow community/committers will be less >> co-operative. >> I just don't think the mono-repo structure model will be sustainable in an >> open source community unless there are long-term vested interests. We can't >> predict that. >> >> The current circular dependency problems between Arrow and Parquet is a >> major problem for the community and it is important. >> >> The current Arrow adaptor code for parquet should live in the arrow repo. >> That will remove a majority of the dependency issues. >> Joshua's work would not have been blocked in parquet-cpp if that adapter >> was in the arrow repo. This will be similar to the ORC adaptor. >> >> The platform API code is pretty stable at this point. Minor changes in the >> future to this code should not be the main reason to combine the arrow >> parquet repos. >> >> " >> *I question whether it's worth the community's time long term to wear* >> >> >> *ourselves out defining custom "ports" / virtual interfaces in eachlibrary >> to plug components together rather than utilizing commonplatform APIs.*" >> >> My answer to your question below would be "Yes". Modularity/separation is >> very important in an open source community where priorities of contributors >> are often short term. >> The retention is low and therefore the acquisition costs should be low as >> well. This is the community over code approach according to me. Minor code >> duplication is not a deal breaker. >> ORC, Parquet, Arrow, etc. are all different components in the big data >> space serving their own functions. >> >> If you still strongly feel
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
A controlled fork doesn’t sound like a terrible option. Copy the code from parquet into arrow, and for a limited period of time it would be the primary. When that period is over, the code in parquet becomes the primary. During the period during which arrow has the primary, the parquet release manager will have to synchronize parquet’s copy of the code (probably by patches) before making releases. Julian > On Jul 31, 2018, at 11:29 AM, Wes McKinney wrote: > >> If you still strongly feel that the only way forward is to clone the >> parquet-cpp repo and part ways, I will withdraw my concern. Having two >> parquet-cpp repos is no way a better approach. > > Yes, indeed. In my view, the next best option after a monorepo is to > fork. That would obviously be a bad outcome for the community. > > It doesn't look like I will be able to convince you that a monorepo is > a good idea; what I would ask instead is that you be willing to give > it a shot, and if it turns out in the way you're describing (which I > don't think it will) then I suggest that we fork at that point. > > - Wes > > On Tue, Jul 31, 2018 at 2:14 PM, Deepak Majeti > wrote: >> Wes, >> >> Unfortunately, I cannot show you any practical fact-based problems of a >> non-existent Arrow-Parquet mono-repo. >> Bringing in related Apache community experiences are more meaningful than >> how mono-repos work at Google and other big organizations. >> We solely depend on volunteers and cannot hire full-time developers. >> You are very well aware of how difficult it has been to find more >> contributors and maintainers for Arrow. parquet-cpp already has a low >> contribution rate to its core components. >> >> We should target to ensure that new volunteers who want to contribute >> bug-fixes/features should spend the least amount of time in figuring out >> the project repo. We can never come up with an automated build system that >> caters to every possible environment. >> My only concern is if the mono-repo will make it harder for new developers >> to work on parquet-cpp core just due to the additional code, build and test >> dependencies. >> I am not saying that the Arrow community/committers will be less >> co-operative. >> I just don't think the mono-repo structure model will be sustainable in an >> open source community unless there are long-term vested interests. We can't >> predict that. >> >> The current circular dependency problems between Arrow and Parquet is a >> major problem for the community and it is important. >> >> The current Arrow adaptor code for parquet should live in the arrow repo. >> That will remove a majority of the dependency issues. >> Joshua's work would not have been blocked in parquet-cpp if that adapter >> was in the arrow repo. This will be similar to the ORC adaptor. >> >> The platform API code is pretty stable at this point. Minor changes in the >> future to this code should not be the main reason to combine the arrow >> parquet repos. >> >> " >> *I question whether it's worth the community's time long term to wear* >> >> >> *ourselves out defining custom "ports" / virtual interfaces in eachlibrary >> to plug components together rather than utilizing commonplatform APIs.*" >> >> My answer to your question below would be "Yes". Modularity/separation is >> very important in an open source community where priorities of contributors >> are often short term. >> The retention is low and therefore the acquisition costs should be low as >> well. This is the community over code approach according to me. Minor code >> duplication is not a deal breaker. >> ORC, Parquet, Arrow, etc. are all different components in the big data >> space serving their own functions. >> >> If you still strongly feel that the only way forward is to clone the >> parquet-cpp repo and part ways, I will withdraw my concern. Having two >> parquet-cpp repos is no way a better approach. >> >> >> >> >> On Tue, Jul 31, 2018 at 10:28 AM Wes McKinney wrote: >> >>> @Antoine >>> By the way, one concern with the monorepo approach: it would slightly >>> increase Arrow CI times (which are already too large). >>> >>> A typical CI run in Arrow is taking about 45 minutes: >>> https://travis-ci.org/apache/arrow/builds/410119750 >>> >>> Parquet run takes about 28 >>> https://travis-ci.org/apache/parquet-cpp/builds/410147208 >>> >>> Inevitably we will need to create some kind of bot to run certain >>> builds on-demand based on commit / PR metadata or on request. >>> >>> The slowest build in Arrow (the Arrow C++/Python one) build could be >>> made substantially shorter by moving some of the slower parts (like >>> the Python ASV benchmarks) from being tested every-commit to nightly >>> or on demand. Using ASAN instead of valgrind in Travis would also >>> improve build times (valgrind build could be moved to a nightly >>> exhaustive test run) >>> >>> - Wes >>> >>> On Mon, Jul 30, 2018 at 10:54 PM, Wes McKinney >>> wrote: > I would like
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
> If you still strongly feel that the only way forward is to clone the > parquet-cpp repo and part ways, I will withdraw my concern. Having two > parquet-cpp repos is no way a better approach. Yes, indeed. In my view, the next best option after a monorepo is to fork. That would obviously be a bad outcome for the community. It doesn't look like I will be able to convince you that a monorepo is a good idea; what I would ask instead is that you be willing to give it a shot, and if it turns out in the way you're describing (which I don't think it will) then I suggest that we fork at that point. - Wes On Tue, Jul 31, 2018 at 2:14 PM, Deepak Majeti wrote: > Wes, > > Unfortunately, I cannot show you any practical fact-based problems of a > non-existent Arrow-Parquet mono-repo. > Bringing in related Apache community experiences are more meaningful than > how mono-repos work at Google and other big organizations. > We solely depend on volunteers and cannot hire full-time developers. > You are very well aware of how difficult it has been to find more > contributors and maintainers for Arrow. parquet-cpp already has a low > contribution rate to its core components. > > We should target to ensure that new volunteers who want to contribute > bug-fixes/features should spend the least amount of time in figuring out > the project repo. We can never come up with an automated build system that > caters to every possible environment. > My only concern is if the mono-repo will make it harder for new developers > to work on parquet-cpp core just due to the additional code, build and test > dependencies. > I am not saying that the Arrow community/committers will be less > co-operative. > I just don't think the mono-repo structure model will be sustainable in an > open source community unless there are long-term vested interests. We can't > predict that. > > The current circular dependency problems between Arrow and Parquet is a > major problem for the community and it is important. > > The current Arrow adaptor code for parquet should live in the arrow repo. > That will remove a majority of the dependency issues. > Joshua's work would not have been blocked in parquet-cpp if that adapter > was in the arrow repo. This will be similar to the ORC adaptor. > > The platform API code is pretty stable at this point. Minor changes in the > future to this code should not be the main reason to combine the arrow > parquet repos. > > " > *I question whether it's worth the community's time long term to wear* > > > *ourselves out defining custom "ports" / virtual interfaces in eachlibrary > to plug components together rather than utilizing commonplatform APIs.*" > > My answer to your question below would be "Yes". Modularity/separation is > very important in an open source community where priorities of contributors > are often short term. > The retention is low and therefore the acquisition costs should be low as > well. This is the community over code approach according to me. Minor code > duplication is not a deal breaker. > ORC, Parquet, Arrow, etc. are all different components in the big data > space serving their own functions. > > If you still strongly feel that the only way forward is to clone the > parquet-cpp repo and part ways, I will withdraw my concern. Having two > parquet-cpp repos is no way a better approach. > > > > > On Tue, Jul 31, 2018 at 10:28 AM Wes McKinney wrote: > >> @Antoine >> >> > By the way, one concern with the monorepo approach: it would slightly >> increase Arrow CI times (which are already too large). >> >> A typical CI run in Arrow is taking about 45 minutes: >> https://travis-ci.org/apache/arrow/builds/410119750 >> >> Parquet run takes about 28 >> https://travis-ci.org/apache/parquet-cpp/builds/410147208 >> >> Inevitably we will need to create some kind of bot to run certain >> builds on-demand based on commit / PR metadata or on request. >> >> The slowest build in Arrow (the Arrow C++/Python one) build could be >> made substantially shorter by moving some of the slower parts (like >> the Python ASV benchmarks) from being tested every-commit to nightly >> or on demand. Using ASAN instead of valgrind in Travis would also >> improve build times (valgrind build could be moved to a nightly >> exhaustive test run) >> >> - Wes >> >> On Mon, Jul 30, 2018 at 10:54 PM, Wes McKinney >> wrote: >> >> I would like to point out that arrow's use of orc is a great example of >> how it would be possible to manage parquet-cpp as a separate codebase. That >> gives me hope that the projects could be managed separately some day. >> > >> > Well, I don't know that ORC is the best example. The ORC C++ codebase >> > features several areas of duplicated logic which could be replaced by >> > components from the Arrow platform for better platform-wide >> > interoperability: >> > >> > >> https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/OrcFile.hh#L37 >> >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Wes, Unfortunately, I cannot show you any practical fact-based problems of a non-existent Arrow-Parquet mono-repo. Bringing in related Apache community experiences are more meaningful than how mono-repos work at Google and other big organizations. We solely depend on volunteers and cannot hire full-time developers. You are very well aware of how difficult it has been to find more contributors and maintainers for Arrow. parquet-cpp already has a low contribution rate to its core components. We should target to ensure that new volunteers who want to contribute bug-fixes/features should spend the least amount of time in figuring out the project repo. We can never come up with an automated build system that caters to every possible environment. My only concern is if the mono-repo will make it harder for new developers to work on parquet-cpp core just due to the additional code, build and test dependencies. I am not saying that the Arrow community/committers will be less co-operative. I just don't think the mono-repo structure model will be sustainable in an open source community unless there are long-term vested interests. We can't predict that. The current circular dependency problems between Arrow and Parquet is a major problem for the community and it is important. The current Arrow adaptor code for parquet should live in the arrow repo. That will remove a majority of the dependency issues. Joshua's work would not have been blocked in parquet-cpp if that adapter was in the arrow repo. This will be similar to the ORC adaptor. The platform API code is pretty stable at this point. Minor changes in the future to this code should not be the main reason to combine the arrow parquet repos. " *I question whether it's worth the community's time long term to wear* *ourselves out defining custom "ports" / virtual interfaces in eachlibrary to plug components together rather than utilizing commonplatform APIs.*" My answer to your question below would be "Yes". Modularity/separation is very important in an open source community where priorities of contributors are often short term. The retention is low and therefore the acquisition costs should be low as well. This is the community over code approach according to me. Minor code duplication is not a deal breaker. ORC, Parquet, Arrow, etc. are all different components in the big data space serving their own functions. If you still strongly feel that the only way forward is to clone the parquet-cpp repo and part ways, I will withdraw my concern. Having two parquet-cpp repos is no way a better approach. On Tue, Jul 31, 2018 at 10:28 AM Wes McKinney wrote: > @Antoine > > > By the way, one concern with the monorepo approach: it would slightly > increase Arrow CI times (which are already too large). > > A typical CI run in Arrow is taking about 45 minutes: > https://travis-ci.org/apache/arrow/builds/410119750 > > Parquet run takes about 28 > https://travis-ci.org/apache/parquet-cpp/builds/410147208 > > Inevitably we will need to create some kind of bot to run certain > builds on-demand based on commit / PR metadata or on request. > > The slowest build in Arrow (the Arrow C++/Python one) build could be > made substantially shorter by moving some of the slower parts (like > the Python ASV benchmarks) from being tested every-commit to nightly > or on demand. Using ASAN instead of valgrind in Travis would also > improve build times (valgrind build could be moved to a nightly > exhaustive test run) > > - Wes > > On Mon, Jul 30, 2018 at 10:54 PM, Wes McKinney > wrote: > >> I would like to point out that arrow's use of orc is a great example of > how it would be possible to manage parquet-cpp as a separate codebase. That > gives me hope that the projects could be managed separately some day. > > > > Well, I don't know that ORC is the best example. The ORC C++ codebase > > features several areas of duplicated logic which could be replaced by > > components from the Arrow platform for better platform-wide > > interoperability: > > > > > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/OrcFile.hh#L37 > > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/Int128.hh > > > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/MemoryPool.hh > > https://github.com/apache/orc/blob/master/c%2B%2B/src/io/InputStream.hh > > https://github.com/apache/orc/blob/master/c%2B%2B/src/io/OutputStream.hh > > > > ORC's use of symbols from Protocol Buffers was actually a cause of > > bugs that we had to fix in Arrow's build system to prevent them from > > leaking to third party linkers when statically linked (ORC is only > > available for static linking at the moment AFAIK). > > > > I question whether it's worth the community's time long term to wear > > ourselves out defining custom "ports" / virtual interfaces in each > > library to plug components together rather than utilizing common > > platform APIs. > > > > - Wes > > > > On Mon, Jul 30, 2018 at 10:45
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
@Antoine > By the way, one concern with the monorepo approach: it would slightly > increase Arrow CI times (which are already too large). A typical CI run in Arrow is taking about 45 minutes: https://travis-ci.org/apache/arrow/builds/410119750 Parquet run takes about 28 https://travis-ci.org/apache/parquet-cpp/builds/410147208 Inevitably we will need to create some kind of bot to run certain builds on-demand based on commit / PR metadata or on request. The slowest build in Arrow (the Arrow C++/Python one) build could be made substantially shorter by moving some of the slower parts (like the Python ASV benchmarks) from being tested every-commit to nightly or on demand. Using ASAN instead of valgrind in Travis would also improve build times (valgrind build could be moved to a nightly exhaustive test run) - Wes On Mon, Jul 30, 2018 at 10:54 PM, Wes McKinney wrote: >> I would like to point out that arrow's use of orc is a great example of how >> it would be possible to manage parquet-cpp as a separate codebase. That >> gives me hope that the projects could be managed separately some day. > > Well, I don't know that ORC is the best example. The ORC C++ codebase > features several areas of duplicated logic which could be replaced by > components from the Arrow platform for better platform-wide > interoperability: > > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/OrcFile.hh#L37 > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/Int128.hh > https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/MemoryPool.hh > https://github.com/apache/orc/blob/master/c%2B%2B/src/io/InputStream.hh > https://github.com/apache/orc/blob/master/c%2B%2B/src/io/OutputStream.hh > > ORC's use of symbols from Protocol Buffers was actually a cause of > bugs that we had to fix in Arrow's build system to prevent them from > leaking to third party linkers when statically linked (ORC is only > available for static linking at the moment AFAIK). > > I question whether it's worth the community's time long term to wear > ourselves out defining custom "ports" / virtual interfaces in each > library to plug components together rather than utilizing common > platform APIs. > > - Wes > > On Mon, Jul 30, 2018 at 10:45 PM, Joshua Storck > wrote: >> You're point about the constraints of the ASF release process are well >> taken and as a developer who's trying to work in the current environment I >> would be much happier if the codebases were merged. The main issues I worry >> about when you put codebases like these together are: >> >> 1. The delineation of API's become blurred and the code becomes too coupled >> 2. Release of artifacts that are lower in the dependency tree are delayed >> by artifacts higher in the dependency tree >> >> If the project/release management is structured well and someone keeps an >> eye on the coupling, then I don't have any concerns. >> >> I would like to point out that arrow's use of orc is a great example of how >> it would be possible to manage parquet-cpp as a separate codebase. That >> gives me hope that the projects could be managed separately some day. >> >> On Mon, Jul 30, 2018 at 10:23 PM Wes McKinney wrote: >> >>> hi Josh, >>> >>> > I can imagine use cases for parquet that don't involve arrow and tying >>> them together seems like the wrong choice. >>> >>> Apache is "Community over Code"; right now it's the same people >>> building these projects -- my argument (which I think you agree with?) >>> is that we should work more closely together until the community grows >>> large enough to support larger-scope process than we have now. As >>> you've seen, our process isn't serving developers of these projects. >>> >>> > I also think build tooling should be pulled into its own codebase. >>> >>> I don't see how this can possibly be practical taking into >>> consideration the constraints imposed by the combination of the GitHub >>> platform and the ASF release process. I'm all for being idealistic, >>> but right now we need to be practical. Unless we can devise a >>> practical procedure that can accommodate at least 1 patch per day >>> which may touch both code and build system simultaneously without >>> being a hindrance to contributor or maintainer, I don't see how we can >>> move forward. >>> >>> > That being said, I think it makes sense to merge the codebases in the >>> short term with the express purpose of separating them in the near term. >>> >>> I would agree but only if separation can be demonstrated to be >>> practical and result in net improvements in productivity and community >>> growth. I think experience has clearly demonstrated that the current >>> separation is impractical, and is causing problems. >>> >>> Per Julian's and Ted's comments, I think we need to consider >>> development process and ASF releases separately. My argument is as >>> follows: >>> >>> * Monorepo for development (for practicality) >>> * Releases structured according to the desires of
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I recently worked on an issue that had to be implemented in parquet-cpp (ARROW-1644, ARROW-1599) but required changes in arrow (ARROW-2585, ARROW-2586). I found the circular dependencies confusing and hard to work with. For example, I still have a PR open in parquet-cpp (created on May 10) because of a PR that it depended on in arrow that was recently merged. I couldn't even address any CI issues in the PR because the change in arrow was not yet in master. In a separate PR, I changed the run_clang_format.py script in the arrow project only to find out later that there was an exact copy of it in parquet-cpp. However, I don't think merging the codebases makes sense in the long term. I can imagine use cases for parquet that don't involve arrow and tying them together seems like the wrong choice. There will be other formats that arrow needs to support that will be kept separate (e.g. - Orc), so I don't see why parquet should be special. I also think build tooling should be pulled into its own codebase. GNU has had a long history of developing open source C/C++ projects that way and made projects like autoconf/automake/make to support them. I don't think CI is a good counter-example since there have been lots of successful open source projects that have used nightly build systems that pinned versions of dependent software. That being said, I think it makes sense to merge the codebases in the short term with the express purpose of separating them in the near term. My reasoning is as follows. By putting the codebases together, you can more easily delineate the boundaries between the API's with a single PR. Second, it will force the build tooling to converge instead of diverge, which has already happened. Once the boundaries and tooling have been sorted out, it should be easy to separate them back into their own codebases. If the codebases are merged, I would ask that the C++ codebases for arrow be separated from other languages. Looking at it from the perspective of a parquet-cpp library user, having a dependency on Java is a large tax to pay if you don't need it. For example, there were 25 JIRA's in the 0.10.0 release of arrow, many of which were holding up the release. I hope that seems like a reasonable compromise, and I think it will help reduce the complexity of the build/release tooling. On Mon, Jul 30, 2018 at 8:50 PM Ted Dunning wrote: > On Mon, Jul 30, 2018 at 5:39 PM Wes McKinney wrote: > > > > > > The community will be less willing to accept large > > > changes that require multiple rounds of patches for stability and API > > > convergence. Our contributions to Libhdfs++ in the HDFS community took > a > > > significantly long time for the very same reason. > > > > Please don't use bad experiences from another open source community as > > leverage in this discussion. I'm sorry that things didn't go the way > > you wanted in Apache Hadoop but this is a distinct community which > > happens to operate under a similar open governance model. > > > There are some more radical and community building options as well. Take > the subversion project as a precedent. With subversion, any Apache > committer can request and receive a commit bit on some large fraction of > subversion. > > So why not take this a bit further and give every parquet committer a > commit bit in Arrow? Or even make them be first class committers in Arrow? > Possibly even make it policy that every Parquet committer who asks will be > given committer status in Arrow. > > That relieves a lot of the social anxiety here. Parquet committers can't be > worried at that point whether their patches will get merged; they can just > merge them. Arrow shouldn't worry much about inviting in the Parquet > committers. After all, Arrow already depends a lot on parquet so why not > invite them in? >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
You're point about the constraints of the ASF release process are well taken and as a developer who's trying to work in the current environment I would be much happier if the codebases were merged. The main issues I worry about when you put codebases like these together are: 1. The delineation of API's become blurred and the code becomes too coupled 2. Release of artifacts that are lower in the dependency tree are delayed by artifacts higher in the dependency tree If the project/release management is structured well and someone keeps an eye on the coupling, then I don't have any concerns. I would like to point out that arrow's use of orc is a great example of how it would be possible to manage parquet-cpp as a separate codebase. That gives me hope that the projects could be managed separately some day. On Mon, Jul 30, 2018 at 10:23 PM Wes McKinney wrote: > hi Josh, > > > I can imagine use cases for parquet that don't involve arrow and tying > them together seems like the wrong choice. > > Apache is "Community over Code"; right now it's the same people > building these projects -- my argument (which I think you agree with?) > is that we should work more closely together until the community grows > large enough to support larger-scope process than we have now. As > you've seen, our process isn't serving developers of these projects. > > > I also think build tooling should be pulled into its own codebase. > > I don't see how this can possibly be practical taking into > consideration the constraints imposed by the combination of the GitHub > platform and the ASF release process. I'm all for being idealistic, > but right now we need to be practical. Unless we can devise a > practical procedure that can accommodate at least 1 patch per day > which may touch both code and build system simultaneously without > being a hindrance to contributor or maintainer, I don't see how we can > move forward. > > > That being said, I think it makes sense to merge the codebases in the > short term with the express purpose of separating them in the near term. > > I would agree but only if separation can be demonstrated to be > practical and result in net improvements in productivity and community > growth. I think experience has clearly demonstrated that the current > separation is impractical, and is causing problems. > > Per Julian's and Ted's comments, I think we need to consider > development process and ASF releases separately. My argument is as > follows: > > * Monorepo for development (for practicality) > * Releases structured according to the desires of the PMCs > > - Wes > > On Mon, Jul 30, 2018 at 9:31 PM, Joshua Storck > wrote: > > I recently worked on an issue that had to be implemented in parquet-cpp > > (ARROW-1644, ARROW-1599) but required changes in arrow (ARROW-2585, > > ARROW-2586). I found the circular dependencies confusing and hard to work > > with. For example, I still have a PR open in parquet-cpp (created on May > > 10) because of a PR that it depended on in arrow that was recently > merged. > > I couldn't even address any CI issues in the PR because the change in > arrow > > was not yet in master. In a separate PR, I changed the > run_clang_format.py > > script in the arrow project only to find out later that there was an > exact > > copy of it in parquet-cpp. > > > > However, I don't think merging the codebases makes sense in the long > term. > > I can imagine use cases for parquet that don't involve arrow and tying > them > > together seems like the wrong choice. There will be other formats that > > arrow needs to support that will be kept separate (e.g. - Orc), so I > don't > > see why parquet should be special. I also think build tooling should be > > pulled into its own codebase. GNU has had a long history of developing > open > > source C/C++ projects that way and made projects like > > autoconf/automake/make to support them. I don't think CI is a good > > counter-example since there have been lots of successful open source > > projects that have used nightly build systems that pinned versions of > > dependent software. > > > > That being said, I think it makes sense to merge the codebases in the > short > > term with the express purpose of separating them in the near term. My > > reasoning is as follows. By putting the codebases together, you can more > > easily delineate the boundaries between the API's with a single PR. > Second, > > it will force the build tooling to converge instead of diverge, which has > > already happened. Once the boundaries and tooling have been sorted out, > it > > should be easy to separate them back into their own codebases. > > > > If the codebases are merged, I would ask that the C++ codebases for arrow > > be separated from other languages. Looking at it from the perspective of > a > > parquet-cpp library user, having a dependency on Java is a large tax to > pay > > if you don't need it. For example, there were 25 JIRA's in the 0.10.0 > > release of arrow,
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
> I would like to point out that arrow's use of orc is a great example of how > it would be possible to manage parquet-cpp as a separate codebase. That gives > me hope that the projects could be managed separately some day. Well, I don't know that ORC is the best example. The ORC C++ codebase features several areas of duplicated logic which could be replaced by components from the Arrow platform for better platform-wide interoperability: https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/OrcFile.hh#L37 https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/Int128.hh https://github.com/apache/orc/blob/master/c%2B%2B/include/orc/MemoryPool.hh https://github.com/apache/orc/blob/master/c%2B%2B/src/io/InputStream.hh https://github.com/apache/orc/blob/master/c%2B%2B/src/io/OutputStream.hh ORC's use of symbols from Protocol Buffers was actually a cause of bugs that we had to fix in Arrow's build system to prevent them from leaking to third party linkers when statically linked (ORC is only available for static linking at the moment AFAIK). I question whether it's worth the community's time long term to wear ourselves out defining custom "ports" / virtual interfaces in each library to plug components together rather than utilizing common platform APIs. - Wes On Mon, Jul 30, 2018 at 10:45 PM, Joshua Storck wrote: > You're point about the constraints of the ASF release process are well > taken and as a developer who's trying to work in the current environment I > would be much happier if the codebases were merged. The main issues I worry > about when you put codebases like these together are: > > 1. The delineation of API's become blurred and the code becomes too coupled > 2. Release of artifacts that are lower in the dependency tree are delayed > by artifacts higher in the dependency tree > > If the project/release management is structured well and someone keeps an > eye on the coupling, then I don't have any concerns. > > I would like to point out that arrow's use of orc is a great example of how > it would be possible to manage parquet-cpp as a separate codebase. That > gives me hope that the projects could be managed separately some day. > > On Mon, Jul 30, 2018 at 10:23 PM Wes McKinney wrote: > >> hi Josh, >> >> > I can imagine use cases for parquet that don't involve arrow and tying >> them together seems like the wrong choice. >> >> Apache is "Community over Code"; right now it's the same people >> building these projects -- my argument (which I think you agree with?) >> is that we should work more closely together until the community grows >> large enough to support larger-scope process than we have now. As >> you've seen, our process isn't serving developers of these projects. >> >> > I also think build tooling should be pulled into its own codebase. >> >> I don't see how this can possibly be practical taking into >> consideration the constraints imposed by the combination of the GitHub >> platform and the ASF release process. I'm all for being idealistic, >> but right now we need to be practical. Unless we can devise a >> practical procedure that can accommodate at least 1 patch per day >> which may touch both code and build system simultaneously without >> being a hindrance to contributor or maintainer, I don't see how we can >> move forward. >> >> > That being said, I think it makes sense to merge the codebases in the >> short term with the express purpose of separating them in the near term. >> >> I would agree but only if separation can be demonstrated to be >> practical and result in net improvements in productivity and community >> growth. I think experience has clearly demonstrated that the current >> separation is impractical, and is causing problems. >> >> Per Julian's and Ted's comments, I think we need to consider >> development process and ASF releases separately. My argument is as >> follows: >> >> * Monorepo for development (for practicality) >> * Releases structured according to the desires of the PMCs >> >> - Wes >> >> On Mon, Jul 30, 2018 at 9:31 PM, Joshua Storck >> wrote: >> > I recently worked on an issue that had to be implemented in parquet-cpp >> > (ARROW-1644, ARROW-1599) but required changes in arrow (ARROW-2585, >> > ARROW-2586). I found the circular dependencies confusing and hard to work >> > with. For example, I still have a PR open in parquet-cpp (created on May >> > 10) because of a PR that it depended on in arrow that was recently >> merged. >> > I couldn't even address any CI issues in the PR because the change in >> arrow >> > was not yet in master. In a separate PR, I changed the >> run_clang_format.py >> > script in the arrow project only to find out later that there was an >> exact >> > copy of it in parquet-cpp. >> > >> > However, I don't think merging the codebases makes sense in the long >> term. >> > I can imagine use cases for parquet that don't involve arrow and tying >> them >> > together seems like the wrong choice. There will be
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi Josh, > I can imagine use cases for parquet that don't involve arrow and tying them > together seems like the wrong choice. Apache is "Community over Code"; right now it's the same people building these projects -- my argument (which I think you agree with?) is that we should work more closely together until the community grows large enough to support larger-scope process than we have now. As you've seen, our process isn't serving developers of these projects. > I also think build tooling should be pulled into its own codebase. I don't see how this can possibly be practical taking into consideration the constraints imposed by the combination of the GitHub platform and the ASF release process. I'm all for being idealistic, but right now we need to be practical. Unless we can devise a practical procedure that can accommodate at least 1 patch per day which may touch both code and build system simultaneously without being a hindrance to contributor or maintainer, I don't see how we can move forward. > That being said, I think it makes sense to merge the codebases in the short > term with the express purpose of separating them in the near term. I would agree but only if separation can be demonstrated to be practical and result in net improvements in productivity and community growth. I think experience has clearly demonstrated that the current separation is impractical, and is causing problems. Per Julian's and Ted's comments, I think we need to consider development process and ASF releases separately. My argument is as follows: * Monorepo for development (for practicality) * Releases structured according to the desires of the PMCs - Wes On Mon, Jul 30, 2018 at 9:31 PM, Joshua Storck wrote: > I recently worked on an issue that had to be implemented in parquet-cpp > (ARROW-1644, ARROW-1599) but required changes in arrow (ARROW-2585, > ARROW-2586). I found the circular dependencies confusing and hard to work > with. For example, I still have a PR open in parquet-cpp (created on May > 10) because of a PR that it depended on in arrow that was recently merged. > I couldn't even address any CI issues in the PR because the change in arrow > was not yet in master. In a separate PR, I changed the run_clang_format.py > script in the arrow project only to find out later that there was an exact > copy of it in parquet-cpp. > > However, I don't think merging the codebases makes sense in the long term. > I can imagine use cases for parquet that don't involve arrow and tying them > together seems like the wrong choice. There will be other formats that > arrow needs to support that will be kept separate (e.g. - Orc), so I don't > see why parquet should be special. I also think build tooling should be > pulled into its own codebase. GNU has had a long history of developing open > source C/C++ projects that way and made projects like > autoconf/automake/make to support them. I don't think CI is a good > counter-example since there have been lots of successful open source > projects that have used nightly build systems that pinned versions of > dependent software. > > That being said, I think it makes sense to merge the codebases in the short > term with the express purpose of separating them in the near term. My > reasoning is as follows. By putting the codebases together, you can more > easily delineate the boundaries between the API's with a single PR. Second, > it will force the build tooling to converge instead of diverge, which has > already happened. Once the boundaries and tooling have been sorted out, it > should be easy to separate them back into their own codebases. > > If the codebases are merged, I would ask that the C++ codebases for arrow > be separated from other languages. Looking at it from the perspective of a > parquet-cpp library user, having a dependency on Java is a large tax to pay > if you don't need it. For example, there were 25 JIRA's in the 0.10.0 > release of arrow, many of which were holding up the release. I hope that > seems like a reasonable compromise, and I think it will help reduce the > complexity of the build/release tooling. > > > On Mon, Jul 30, 2018 at 8:50 PM Ted Dunning wrote: > >> On Mon, Jul 30, 2018 at 5:39 PM Wes McKinney wrote: >> >> > >> > > The community will be less willing to accept large >> > > changes that require multiple rounds of patches for stability and API >> > > convergence. Our contributions to Libhdfs++ in the HDFS community took >> a >> > > significantly long time for the very same reason. >> > >> > Please don't use bad experiences from another open source community as >> > leverage in this discussion. I'm sorry that things didn't go the way >> > you wanted in Apache Hadoop but this is a distinct community which >> > happens to operate under a similar open governance model. >> >> >> There are some more radical and community building options as well. Take >> the subversion project as a precedent. With subversion, any Apache >> committer can
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
On Mon, Jul 30, 2018 at 8:50 PM, Ted Dunning wrote: > On Mon, Jul 30, 2018 at 5:39 PM Wes McKinney wrote: > >> >> > The community will be less willing to accept large >> > changes that require multiple rounds of patches for stability and API >> > convergence. Our contributions to Libhdfs++ in the HDFS community took a >> > significantly long time for the very same reason. >> >> Please don't use bad experiences from another open source community as >> leverage in this discussion. I'm sorry that things didn't go the way >> you wanted in Apache Hadoop but this is a distinct community which >> happens to operate under a similar open governance model. > > > There are some more radical and community building options as well. Take > the subversion project as a precedent. With subversion, any Apache > committer can request and receive a commit bit on some large fraction of > subversion. > > So why not take this a bit further and give every parquet committer a > commit bit in Arrow? Or even make them be first class committers in Arrow? > Possibly even make it policy that every Parquet committer who asks will be > given committer status in Arrow. > > That relieves a lot of the social anxiety here. Parquet committers can't be > worried at that point whether their patches will get merged; they can just > merge them. Arrow shouldn't worry much about inviting in the Parquet > committers. After all, Arrow already depends a lot on parquet so why not > invite them in? hi Ted, I for one am with you on this idea, and don't see it as all that radical. The Arrow and Parquet communities are working toward the same goals: open standards for storage and in-memory analytics. This is part of why so there is so much overlap already amongst the committers and PMC members. We are stronger working together than fragmented. - Wes
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
On Mon, Jul 30, 2018 at 5:39 PM Wes McKinney wrote: > > > The community will be less willing to accept large > > changes that require multiple rounds of patches for stability and API > > convergence. Our contributions to Libhdfs++ in the HDFS community took a > > significantly long time for the very same reason. > > Please don't use bad experiences from another open source community as > leverage in this discussion. I'm sorry that things didn't go the way > you wanted in Apache Hadoop but this is a distinct community which > happens to operate under a similar open governance model. There are some more radical and community building options as well. Take the subversion project as a precedent. With subversion, any Apache committer can request and receive a commit bit on some large fraction of subversion. So why not take this a bit further and give every parquet committer a commit bit in Arrow? Or even make them be first class committers in Arrow? Possibly even make it policy that every Parquet committer who asks will be given committer status in Arrow. That relieves a lot of the social anxiety here. Parquet committers can't be worried at that point whether their patches will get merged; they can just merge them. Arrow shouldn't worry much about inviting in the Parquet committers. After all, Arrow already depends a lot on parquet so why not invite them in?
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi, On Mon, Jul 30, 2018 at 6:52 PM, Deepak Majeti wrote: > Wes, > > I definitely appreciate and do see the impact of contributions made by > everyone. I made this statement not to rate any contributions but solely to > support my concern. > The contribution barrier is higher simply because of the increased code, > build, and test dependencies. If the community has lesser interest on a > certain component (parquet-cpp core in this case), it becomes very hard to > make big changes. This is a FUD-based argument rather than a fact-based one. If there are committers in Arrow (via Parquet) who approve changes, why would they not be merged? The community will be incentivized to make sure that developers are productive and able to work efficiently on the part of the project that is relevant to them. parquet-cpp developers are already building most of Arrow's C++ codebase en route to development; by building with a single build system development environments will be simpler to manage in general. On the subject of code velocity: Arrow has a diverse community and nearly 200 unique contributors at this point. Parquet has about 30. The Arrow codebase history starts on February 5, 2016. Since then: * Arrow has has 2055 patches * parquet-cpp has had 425 So the patch volume is about 5x as high on average. This does not look like a project that is struggling to merge patches. We are all invested in the success of Parquet and changing the structure of the code and helping the community to work more productively would not change that. > The community will be less willing to accept large > changes that require multiple rounds of patches for stability and API > convergence. Our contributions to Libhdfs++ in the HDFS community took a > significantly long time for the very same reason. Please don't use bad experiences from another open source community as leverage in this discussion. I'm sorry that things didn't go the way you wanted in Apache Hadoop but this is a distinct community which happens to operate under a similar open governance model. After significant time thinking about it, I think unfortunately that the next-best option after a monorepo structure would be for the Arrow community to _fork_ the parquet-cpp codebase and go our separate ways. Beyond these two options I fail to see a pragmatic solution to the problems we've been having. - Wes > > > > On Mon, Jul 30, 2018 at 6:05 PM Wes McKinney wrote: > >> hi Deepak >> >> On Mon, Jul 30, 2018 at 5:18 PM, Deepak Majeti >> wrote: >> > @Wes >> > My observation is that most of the parquet-cpp contributors you listed >> that >> > overlap with the Arrow community mainly contribute to the Arrow >> > bindings(parquet::arrow layer)/platform API changes in the parquet-cpp >> > repo. Very few of them review/contribute patches to the parquet-cpp core. >> > >> >> So, what are you saying exactly, that some contributions or >> contributors to Apache Parquet matter more than others? I don't >> follow. >> >> As a result of these individual's efforts, the parquet-cpp libraries >> are being installed well over 100,000 times per month on a single >> install path (Python) alone. >> > >> > I believe improvements to the parquet-cpp core will be negatively >> impacted >> > since merging the parquet-cpp and arrow-cpp repos will increase the >> barrier >> > of entry to new contributors interested in the parquet-cpp core. The >> > current extensions to the parquet-cpp core related to bloom-filters, and >> > column encryption are all being done by first-time contributors. >> >> I don't understand why this would "increase the barrier of entry". >> Could you explain? >> > It is true that there would be more code in the codebase, but the >> build and test procedure would be no more complex. If anything, >> community productivity will be improved by having a more cohesive / >> centralized development platform (large amounts of code that Parquet >> depends on are in Apache Arrow already). >> >> > >> > If you believe there will be new interest in the parquet-cpp core with >> the >> > mono-repo approach, I am all up for it. >> >> Yes, I believe that this change will result in more and higher quality >> code review to Parquet core changes and general improvements to >> developer productivity across the board. Developer productivity is >> what this is all about. >> >> - Wes >> >> > >> > >> > On Mon, Jul 30, 2018 at 12:18 AM Philipp Moritz >> wrote: >> > >> >> I do not claim to have insight into parquet-cpp development. However, >> from >> >> our experience developing Ray, I can say that the monorepo approach (for >> >> Ray) has improved things a lot. Before we tried various schemes to split >> >> the project into multiple repos, but the build system and test >> >> infrastructure duplications and overhead from synchronizing changes >> slowed >> >> development down significantly (and fixing bugs that touch the subrepos >> and >> >> the main repo is inconvenient). >> >> >> >> Also the
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I'm not going to comment on the design of the parquet-cpp module and whether it is “closer” to parquet or arrow. But I do think Wes’s proposal is consistent with Apache policy. PMCs make releases and govern communities; they don’t exist to manage code bases, except as a means to the end of creating releases of known provenance. The Parquet PMC can continue to make parquet-cpp releases, and to end-users those releases will look the same as they do today, even if the code for those releases were to move to a different git repo in the ASF. Julian > On Jul 30, 2018, at 3:05 PM, Wes McKinney wrote: > > hi Deepak > > On Mon, Jul 30, 2018 at 5:18 PM, Deepak Majeti > wrote: >> @Wes >> My observation is that most of the parquet-cpp contributors you listed that >> overlap with the Arrow community mainly contribute to the Arrow >> bindings(parquet::arrow layer)/platform API changes in the parquet-cpp >> repo. Very few of them review/contribute patches to the parquet-cpp core. >> > > So, what are you saying exactly, that some contributions or > contributors to Apache Parquet matter more than others? I don't > follow. > > As a result of these individual's efforts, the parquet-cpp libraries > are being installed well over 100,000 times per month on a single > install path (Python) alone. > >> I believe improvements to the parquet-cpp core will be negatively impacted >> since merging the parquet-cpp and arrow-cpp repos will increase the barrier >> of entry to new contributors interested in the parquet-cpp core. The >> current extensions to the parquet-cpp core related to bloom-filters, and >> column encryption are all being done by first-time contributors. > > I don't understand why this would "increase the barrier of entry". > Could you explain? > > It is true that there would be more code in the codebase, but the > build and test procedure would be no more complex. If anything, > community productivity will be improved by having a more cohesive / > centralized development platform (large amounts of code that Parquet > depends on are in Apache Arrow already). > >> >> If you believe there will be new interest in the parquet-cpp core with the >> mono-repo approach, I am all up for it. > > Yes, I believe that this change will result in more and higher quality > code review to Parquet core changes and general improvements to > developer productivity across the board. Developer productivity is > what this is all about. > > - Wes > >> >> >> On Mon, Jul 30, 2018 at 12:18 AM Philipp Moritz wrote: >> >>> I do not claim to have insight into parquet-cpp development. However, from >>> our experience developing Ray, I can say that the monorepo approach (for >>> Ray) has improved things a lot. Before we tried various schemes to split >>> the project into multiple repos, but the build system and test >>> infrastructure duplications and overhead from synchronizing changes slowed >>> development down significantly (and fixing bugs that touch the subrepos and >>> the main repo is inconvenient). >>> >>> Also the decision to put arrow and parquet-cpp into a common repo is >>> independent of how tightly coupled the two projects are (and there could be >>> a matrix entry in travis which tests that PRs keep them decoupled, or >>> rather that they both just depend on a small common "base"). Google and >>> Facebook demonstrate such independence by having many many projects in the >>> same repo of course. It would be great if the open source community would >>> move more into this direction too I think. >>> >>> Best, >>> Philipp. >>> >>> On Sun, Jul 29, 2018 at 8:54 PM, Wes McKinney wrote: >>> hi Donald, This would make things worse, not better. Code changes routinely involve changes to the build system, and so you could be talking about having to making changes to 2 or 3 git repositories as the result of a single new feature or bug fix. There isn't really a cross-repo CI solution available I've seen some approaches to the monorepo problem using multiple git repositories, such as https://github.com/twosigma/git-meta Until something like this has first class support by the GitHub platform and its CI services (Travis CI, Appveyor), I don't think it will work for us. - Wes On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss wrote: > Could this work as each module gets configured as sub-git repots. Top level > build tool go into each sub-repo, pick the correct release version to test. > Tests in Python is dependent on cpp sub-repo to ensure the API still pass. > > This should be the best of both worlds, if sub-repo are supposed >>> option. > > --Donald E. Foss > > On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti > wrote: > >> I dislike the current build system complications as well. >> >> However, in my opinion, combining the code
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Wes, I definitely appreciate and do see the impact of contributions made by everyone. I made this statement not to rate any contributions but solely to support my concern. The contribution barrier is higher simply because of the increased code, build, and test dependencies. If the community has lesser interest on a certain component (parquet-cpp core in this case), it becomes very hard to make big changes. The community will be less willing to accept large changes that require multiple rounds of patches for stability and API convergence. Our contributions to Libhdfs++ in the HDFS community took a significantly long time for the very same reason. On Mon, Jul 30, 2018 at 6:05 PM Wes McKinney wrote: > hi Deepak > > On Mon, Jul 30, 2018 at 5:18 PM, Deepak Majeti > wrote: > > @Wes > > My observation is that most of the parquet-cpp contributors you listed > that > > overlap with the Arrow community mainly contribute to the Arrow > > bindings(parquet::arrow layer)/platform API changes in the parquet-cpp > > repo. Very few of them review/contribute patches to the parquet-cpp core. > > > > So, what are you saying exactly, that some contributions or > contributors to Apache Parquet matter more than others? I don't > follow. > > As a result of these individual's efforts, the parquet-cpp libraries > are being installed well over 100,000 times per month on a single > install path (Python) alone. > > > I believe improvements to the parquet-cpp core will be negatively > impacted > > since merging the parquet-cpp and arrow-cpp repos will increase the > barrier > > of entry to new contributors interested in the parquet-cpp core. The > > current extensions to the parquet-cpp core related to bloom-filters, and > > column encryption are all being done by first-time contributors. > > I don't understand why this would "increase the barrier of entry". > Could you explain? > It is true that there would be more code in the codebase, but the > build and test procedure would be no more complex. If anything, > community productivity will be improved by having a more cohesive / > centralized development platform (large amounts of code that Parquet > depends on are in Apache Arrow already). > > > > > If you believe there will be new interest in the parquet-cpp core with > the > > mono-repo approach, I am all up for it. > > Yes, I believe that this change will result in more and higher quality > code review to Parquet core changes and general improvements to > developer productivity across the board. Developer productivity is > what this is all about. > > - Wes > > > > > > > On Mon, Jul 30, 2018 at 12:18 AM Philipp Moritz > wrote: > > > >> I do not claim to have insight into parquet-cpp development. However, > from > >> our experience developing Ray, I can say that the monorepo approach (for > >> Ray) has improved things a lot. Before we tried various schemes to split > >> the project into multiple repos, but the build system and test > >> infrastructure duplications and overhead from synchronizing changes > slowed > >> development down significantly (and fixing bugs that touch the subrepos > and > >> the main repo is inconvenient). > >> > >> Also the decision to put arrow and parquet-cpp into a common repo is > >> independent of how tightly coupled the two projects are (and there > could be > >> a matrix entry in travis which tests that PRs keep them decoupled, or > >> rather that they both just depend on a small common "base"). Google and > >> Facebook demonstrate such independence by having many many projects in > the > >> same repo of course. It would be great if the open source community > would > >> move more into this direction too I think. > >> > >> Best, > >> Philipp. > >> > >> On Sun, Jul 29, 2018 at 8:54 PM, Wes McKinney > wrote: > >> > >> > hi Donald, > >> > > >> > This would make things worse, not better. Code changes routinely > >> > involve changes to the build system, and so you could be talking about > >> > having to making changes to 2 or 3 git repositories as the result of a > >> > single new feature or bug fix. There isn't really a cross-repo CI > >> > solution available > >> > > >> > I've seen some approaches to the monorepo problem using multiple git > >> > repositories, such as > >> > > >> > https://github.com/twosigma/git-meta > >> > > >> > Until something like this has first class support by the GitHub > >> > platform and its CI services (Travis CI, Appveyor), I don't think it > >> > will work for us. > >> > > >> > - Wes > >> > > >> > On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss < > donald.f...@gmail.com> > >> > wrote: > >> > > Could this work as each module gets configured as sub-git repots. > Top > >> > level > >> > > build tool go into each sub-repo, pick the correct release version > to > >> > test. > >> > > Tests in Python is dependent on cpp sub-repo to ensure the API still > >> > pass. > >> > > > >> > > This should be the best of both worlds, if sub-repo are supposed > >> option. > >> > > >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi Deepak On Mon, Jul 30, 2018 at 5:18 PM, Deepak Majeti wrote: > @Wes > My observation is that most of the parquet-cpp contributors you listed that > overlap with the Arrow community mainly contribute to the Arrow > bindings(parquet::arrow layer)/platform API changes in the parquet-cpp > repo. Very few of them review/contribute patches to the parquet-cpp core. > So, what are you saying exactly, that some contributions or contributors to Apache Parquet matter more than others? I don't follow. As a result of these individual's efforts, the parquet-cpp libraries are being installed well over 100,000 times per month on a single install path (Python) alone. > I believe improvements to the parquet-cpp core will be negatively impacted > since merging the parquet-cpp and arrow-cpp repos will increase the barrier > of entry to new contributors interested in the parquet-cpp core. The > current extensions to the parquet-cpp core related to bloom-filters, and > column encryption are all being done by first-time contributors. I don't understand why this would "increase the barrier of entry". Could you explain? It is true that there would be more code in the codebase, but the build and test procedure would be no more complex. If anything, community productivity will be improved by having a more cohesive / centralized development platform (large amounts of code that Parquet depends on are in Apache Arrow already). > > If you believe there will be new interest in the parquet-cpp core with the > mono-repo approach, I am all up for it. Yes, I believe that this change will result in more and higher quality code review to Parquet core changes and general improvements to developer productivity across the board. Developer productivity is what this is all about. - Wes > > > On Mon, Jul 30, 2018 at 12:18 AM Philipp Moritz wrote: > >> I do not claim to have insight into parquet-cpp development. However, from >> our experience developing Ray, I can say that the monorepo approach (for >> Ray) has improved things a lot. Before we tried various schemes to split >> the project into multiple repos, but the build system and test >> infrastructure duplications and overhead from synchronizing changes slowed >> development down significantly (and fixing bugs that touch the subrepos and >> the main repo is inconvenient). >> >> Also the decision to put arrow and parquet-cpp into a common repo is >> independent of how tightly coupled the two projects are (and there could be >> a matrix entry in travis which tests that PRs keep them decoupled, or >> rather that they both just depend on a small common "base"). Google and >> Facebook demonstrate such independence by having many many projects in the >> same repo of course. It would be great if the open source community would >> move more into this direction too I think. >> >> Best, >> Philipp. >> >> On Sun, Jul 29, 2018 at 8:54 PM, Wes McKinney wrote: >> >> > hi Donald, >> > >> > This would make things worse, not better. Code changes routinely >> > involve changes to the build system, and so you could be talking about >> > having to making changes to 2 or 3 git repositories as the result of a >> > single new feature or bug fix. There isn't really a cross-repo CI >> > solution available >> > >> > I've seen some approaches to the monorepo problem using multiple git >> > repositories, such as >> > >> > https://github.com/twosigma/git-meta >> > >> > Until something like this has first class support by the GitHub >> > platform and its CI services (Travis CI, Appveyor), I don't think it >> > will work for us. >> > >> > - Wes >> > >> > On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss >> > wrote: >> > > Could this work as each module gets configured as sub-git repots. Top >> > level >> > > build tool go into each sub-repo, pick the correct release version to >> > test. >> > > Tests in Python is dependent on cpp sub-repo to ensure the API still >> > pass. >> > > >> > > This should be the best of both worlds, if sub-repo are supposed >> option. >> > > >> > > --Donald E. Foss >> > > >> > > On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti >> > > wrote: >> > > >> > >> I dislike the current build system complications as well. >> > >> >> > >> However, in my opinion, combining the code bases will severely impact >> > the >> > >> progress of the parquet-cpp project and implicitly the progress of the >> > >> entire parquet project. >> > >> Combining would have made much more sense if parquet-cpp is a mature >> > >> project and codebase. But parquet-cpp (and the entire parquet >> project) >> > is >> > >> evolving continuously with new features being added including bloom >> > >> filters, column encryption, and indexes. >> > >> >> > >> If the two code bases merged, it will be much more difficult to >> > contribute >> > >> to the parquet-cpp project since now Arrow bindings have to be >> > supported as >> > >> well. Please correct me if I am wrong here. >> > >> >> > >> Out of the two evils, I think
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
@Wes My observation is that most of the parquet-cpp contributors you listed that overlap with the Arrow community mainly contribute to the Arrow bindings(parquet::arrow layer)/platform API changes in the parquet-cpp repo. Very few of them review/contribute patches to the parquet-cpp core. I believe improvements to the parquet-cpp core will be negatively impacted since merging the parquet-cpp and arrow-cpp repos will increase the barrier of entry to new contributors interested in the parquet-cpp core. The current extensions to the parquet-cpp core related to bloom-filters, and column encryption are all being done by first-time contributors. If you believe there will be new interest in the parquet-cpp core with the mono-repo approach, I am all up for it. On Mon, Jul 30, 2018 at 12:18 AM Philipp Moritz wrote: > I do not claim to have insight into parquet-cpp development. However, from > our experience developing Ray, I can say that the monorepo approach (for > Ray) has improved things a lot. Before we tried various schemes to split > the project into multiple repos, but the build system and test > infrastructure duplications and overhead from synchronizing changes slowed > development down significantly (and fixing bugs that touch the subrepos and > the main repo is inconvenient). > > Also the decision to put arrow and parquet-cpp into a common repo is > independent of how tightly coupled the two projects are (and there could be > a matrix entry in travis which tests that PRs keep them decoupled, or > rather that they both just depend on a small common "base"). Google and > Facebook demonstrate such independence by having many many projects in the > same repo of course. It would be great if the open source community would > move more into this direction too I think. > > Best, > Philipp. > > On Sun, Jul 29, 2018 at 8:54 PM, Wes McKinney wrote: > > > hi Donald, > > > > This would make things worse, not better. Code changes routinely > > involve changes to the build system, and so you could be talking about > > having to making changes to 2 or 3 git repositories as the result of a > > single new feature or bug fix. There isn't really a cross-repo CI > > solution available > > > > I've seen some approaches to the monorepo problem using multiple git > > repositories, such as > > > > https://github.com/twosigma/git-meta > > > > Until something like this has first class support by the GitHub > > platform and its CI services (Travis CI, Appveyor), I don't think it > > will work for us. > > > > - Wes > > > > On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss > > wrote: > > > Could this work as each module gets configured as sub-git repots. Top > > level > > > build tool go into each sub-repo, pick the correct release version to > > test. > > > Tests in Python is dependent on cpp sub-repo to ensure the API still > > pass. > > > > > > This should be the best of both worlds, if sub-repo are supposed > option. > > > > > > --Donald E. Foss > > > > > > On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti > > > wrote: > > > > > >> I dislike the current build system complications as well. > > >> > > >> However, in my opinion, combining the code bases will severely impact > > the > > >> progress of the parquet-cpp project and implicitly the progress of the > > >> entire parquet project. > > >> Combining would have made much more sense if parquet-cpp is a mature > > >> project and codebase. But parquet-cpp (and the entire parquet > project) > > is > > >> evolving continuously with new features being added including bloom > > >> filters, column encryption, and indexes. > > >> > > >> If the two code bases merged, it will be much more difficult to > > contribute > > >> to the parquet-cpp project since now Arrow bindings have to be > > supported as > > >> well. Please correct me if I am wrong here. > > >> > > >> Out of the two evils, I think handling the build system, packaging > > >> duplication is much more manageable since they are quite stable at > this > > >> point. > > >> > > >> Regarding "* API changes cause awkward release coordination issues > > between > > >> Arrow and Parquet". Can we make minor releases for parquet-cpp (with > API > > >> changes needed) as and when Arrow is released? > > >> > > >> Regarding "we maintain a Arrow conversion code in parquet-cpp for > > >> converting between Arrow columnar memory format and Parquet". Can this > > be > > >> moved to the Arrow project and expose the more stable low-level APIs > in > > >> parquet-cpp? > > >> > > >> I am also curious if the Arrow and Parquet Java implementations have > > >> similar API compatibility issues. > > >> > > >> > > >> On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney > > wrote: > > >> > > >> > hi folks, > > >> > > > >> > We've been struggling for quite some time with the development > > >> > workflow between the Arrow and Parquet C++ (and Python) codebases. > > >> > > > >> > To explain the root issues: > > >> > > > >> > * parquet-cpp depends on
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
Could this work as each module gets configured as sub-git repots. Top level build tool go into each sub-repo, pick the correct release version to test. Tests in Python is dependent on cpp sub-repo to ensure the API still pass. This should be the best of both worlds, if sub-repo are supposed option. --Donald E. Foss On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti wrote: > I dislike the current build system complications as well. > > However, in my opinion, combining the code bases will severely impact the > progress of the parquet-cpp project and implicitly the progress of the > entire parquet project. > Combining would have made much more sense if parquet-cpp is a mature > project and codebase. But parquet-cpp (and the entire parquet project) is > evolving continuously with new features being added including bloom > filters, column encryption, and indexes. > > If the two code bases merged, it will be much more difficult to contribute > to the parquet-cpp project since now Arrow bindings have to be supported as > well. Please correct me if I am wrong here. > > Out of the two evils, I think handling the build system, packaging > duplication is much more manageable since they are quite stable at this > point. > > Regarding "* API changes cause awkward release coordination issues between > Arrow and Parquet". Can we make minor releases for parquet-cpp (with API > changes needed) as and when Arrow is released? > > Regarding "we maintain a Arrow conversion code in parquet-cpp for > converting between Arrow columnar memory format and Parquet". Can this be > moved to the Arrow project and expose the more stable low-level APIs in > parquet-cpp? > > I am also curious if the Arrow and Parquet Java implementations have > similar API compatibility issues. > > > On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney wrote: > > > hi folks, > > > > We've been struggling for quite some time with the development > > workflow between the Arrow and Parquet C++ (and Python) codebases. > > > > To explain the root issues: > > > > * parquet-cpp depends on "platform code" in Apache Arrow; this > > includes file interfaces, memory management, miscellaneous algorithms > > (e.g. dictionary encoding), etc. Note that before this "platform" > > dependency was introduced, there was significant duplicated code > > between these codebases and incompatible abstract interfaces for > > things like files > > > > * we maintain a Arrow conversion code in parquet-cpp for converting > > between Arrow columnar memory format and Parquet > > > > * we maintain Python bindings for parquet-cpp + Arrow interop in > > Apache Arrow. This introduces a circular dependency into our CI. > > > > * Substantial portions of our CMake build system and related tooling > > are duplicated between the Arrow and Parquet repos > > > > * API changes cause awkward release coordination issues between Arrow > > and Parquet > > > > I believe the best way to remedy the situation is to adopt a > > "Community over Code" approach and find a way for the Parquet and > > Arrow C++ development communities to operate out of the same code > > repository, i.e. the apache/arrow git repository. > > > > This would bring major benefits: > > > > * Shared CMake build infrastructure, developer tools, and CI > > infrastructure (Parquet is already being built as a dependency in > > Arrow's CI systems) > > > > * Share packaging and release management infrastructure > > > > * Reduce / eliminate problems due to API changes (where we currently > > introduce breakage into our CI workflow when there is a breaking / > > incompatible change) > > > > * Arrow releases would include a coordinated snapshot of the Parquet > > implementation as it stands > > > > Continuing with the status quo has become unsatisfactory to me and as > > a result I've become less motivated to work on the parquet-cpp > > codebase. > > > > The only Parquet C++ committer who is not an Arrow committer is Deepak > > Majeti. I think the issue of commit privileges could be resolved > > without too much difficulty or time. > > > > I also think if it is truly necessary that the Apache Parquet > > community could create release scripts to cut a miniml versioned > > Apache Parquet C++ release if that is deemed truly necessary. > > > > I know that some people are wary of monorepos and megaprojects, but as > > an example TensorFlow is at least 10 times as large of a projects in > > terms of LOCs and number of different platform components, and it > > seems to be getting along just fine. I think we should be able to work > > together as a community to function just as well. > > > > Interested in the opinions of others, and any other ideas for > > practical solutions to the above problems. > > > > Thanks, > > Wes > > > > > -- > regards, > Deepak Majeti >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I do not claim to have insight into parquet-cpp development. However, from our experience developing Ray, I can say that the monorepo approach (for Ray) has improved things a lot. Before we tried various schemes to split the project into multiple repos, but the build system and test infrastructure duplications and overhead from synchronizing changes slowed development down significantly (and fixing bugs that touch the subrepos and the main repo is inconvenient). Also the decision to put arrow and parquet-cpp into a common repo is independent of how tightly coupled the two projects are (and there could be a matrix entry in travis which tests that PRs keep them decoupled, or rather that they both just depend on a small common "base"). Google and Facebook demonstrate such independence by having many many projects in the same repo of course. It would be great if the open source community would move more into this direction too I think. Best, Philipp. On Sun, Jul 29, 2018 at 8:54 PM, Wes McKinney wrote: > hi Donald, > > This would make things worse, not better. Code changes routinely > involve changes to the build system, and so you could be talking about > having to making changes to 2 or 3 git repositories as the result of a > single new feature or bug fix. There isn't really a cross-repo CI > solution available > > I've seen some approaches to the monorepo problem using multiple git > repositories, such as > > https://github.com/twosigma/git-meta > > Until something like this has first class support by the GitHub > platform and its CI services (Travis CI, Appveyor), I don't think it > will work for us. > > - Wes > > On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss > wrote: > > Could this work as each module gets configured as sub-git repots. Top > level > > build tool go into each sub-repo, pick the correct release version to > test. > > Tests in Python is dependent on cpp sub-repo to ensure the API still > pass. > > > > This should be the best of both worlds, if sub-repo are supposed option. > > > > --Donald E. Foss > > > > On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti > > wrote: > > > >> I dislike the current build system complications as well. > >> > >> However, in my opinion, combining the code bases will severely impact > the > >> progress of the parquet-cpp project and implicitly the progress of the > >> entire parquet project. > >> Combining would have made much more sense if parquet-cpp is a mature > >> project and codebase. But parquet-cpp (and the entire parquet project) > is > >> evolving continuously with new features being added including bloom > >> filters, column encryption, and indexes. > >> > >> If the two code bases merged, it will be much more difficult to > contribute > >> to the parquet-cpp project since now Arrow bindings have to be > supported as > >> well. Please correct me if I am wrong here. > >> > >> Out of the two evils, I think handling the build system, packaging > >> duplication is much more manageable since they are quite stable at this > >> point. > >> > >> Regarding "* API changes cause awkward release coordination issues > between > >> Arrow and Parquet". Can we make minor releases for parquet-cpp (with API > >> changes needed) as and when Arrow is released? > >> > >> Regarding "we maintain a Arrow conversion code in parquet-cpp for > >> converting between Arrow columnar memory format and Parquet". Can this > be > >> moved to the Arrow project and expose the more stable low-level APIs in > >> parquet-cpp? > >> > >> I am also curious if the Arrow and Parquet Java implementations have > >> similar API compatibility issues. > >> > >> > >> On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney > wrote: > >> > >> > hi folks, > >> > > >> > We've been struggling for quite some time with the development > >> > workflow between the Arrow and Parquet C++ (and Python) codebases. > >> > > >> > To explain the root issues: > >> > > >> > * parquet-cpp depends on "platform code" in Apache Arrow; this > >> > includes file interfaces, memory management, miscellaneous algorithms > >> > (e.g. dictionary encoding), etc. Note that before this "platform" > >> > dependency was introduced, there was significant duplicated code > >> > between these codebases and incompatible abstract interfaces for > >> > things like files > >> > > >> > * we maintain a Arrow conversion code in parquet-cpp for converting > >> > between Arrow columnar memory format and Parquet > >> > > >> > * we maintain Python bindings for parquet-cpp + Arrow interop in > >> > Apache Arrow. This introduces a circular dependency into our CI. > >> > > >> > * Substantial portions of our CMake build system and related tooling > >> > are duplicated between the Arrow and Parquet repos > >> > > >> > * API changes cause awkward release coordination issues between Arrow > >> > and Parquet > >> > > >> > I believe the best way to remedy the situation is to adopt a > >> > "Community over Code" approach and find a way for the Parquet and > >> >
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi Donald, This would make things worse, not better. Code changes routinely involve changes to the build system, and so you could be talking about having to making changes to 2 or 3 git repositories as the result of a single new feature or bug fix. There isn't really a cross-repo CI solution available I've seen some approaches to the monorepo problem using multiple git repositories, such as https://github.com/twosigma/git-meta Until something like this has first class support by the GitHub platform and its CI services (Travis CI, Appveyor), I don't think it will work for us. - Wes On Sun, Jul 29, 2018 at 10:54 PM, Donald E. Foss wrote: > Could this work as each module gets configured as sub-git repots. Top level > build tool go into each sub-repo, pick the correct release version to test. > Tests in Python is dependent on cpp sub-repo to ensure the API still pass. > > This should be the best of both worlds, if sub-repo are supposed option. > > --Donald E. Foss > > On Sun, Jul 29, 2018, 10:44 PM Deepak Majeti > wrote: > >> I dislike the current build system complications as well. >> >> However, in my opinion, combining the code bases will severely impact the >> progress of the parquet-cpp project and implicitly the progress of the >> entire parquet project. >> Combining would have made much more sense if parquet-cpp is a mature >> project and codebase. But parquet-cpp (and the entire parquet project) is >> evolving continuously with new features being added including bloom >> filters, column encryption, and indexes. >> >> If the two code bases merged, it will be much more difficult to contribute >> to the parquet-cpp project since now Arrow bindings have to be supported as >> well. Please correct me if I am wrong here. >> >> Out of the two evils, I think handling the build system, packaging >> duplication is much more manageable since they are quite stable at this >> point. >> >> Regarding "* API changes cause awkward release coordination issues between >> Arrow and Parquet". Can we make minor releases for parquet-cpp (with API >> changes needed) as and when Arrow is released? >> >> Regarding "we maintain a Arrow conversion code in parquet-cpp for >> converting between Arrow columnar memory format and Parquet". Can this be >> moved to the Arrow project and expose the more stable low-level APIs in >> parquet-cpp? >> >> I am also curious if the Arrow and Parquet Java implementations have >> similar API compatibility issues. >> >> >> On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney wrote: >> >> > hi folks, >> > >> > We've been struggling for quite some time with the development >> > workflow between the Arrow and Parquet C++ (and Python) codebases. >> > >> > To explain the root issues: >> > >> > * parquet-cpp depends on "platform code" in Apache Arrow; this >> > includes file interfaces, memory management, miscellaneous algorithms >> > (e.g. dictionary encoding), etc. Note that before this "platform" >> > dependency was introduced, there was significant duplicated code >> > between these codebases and incompatible abstract interfaces for >> > things like files >> > >> > * we maintain a Arrow conversion code in parquet-cpp for converting >> > between Arrow columnar memory format and Parquet >> > >> > * we maintain Python bindings for parquet-cpp + Arrow interop in >> > Apache Arrow. This introduces a circular dependency into our CI. >> > >> > * Substantial portions of our CMake build system and related tooling >> > are duplicated between the Arrow and Parquet repos >> > >> > * API changes cause awkward release coordination issues between Arrow >> > and Parquet >> > >> > I believe the best way to remedy the situation is to adopt a >> > "Community over Code" approach and find a way for the Parquet and >> > Arrow C++ development communities to operate out of the same code >> > repository, i.e. the apache/arrow git repository. >> > >> > This would bring major benefits: >> > >> > * Shared CMake build infrastructure, developer tools, and CI >> > infrastructure (Parquet is already being built as a dependency in >> > Arrow's CI systems) >> > >> > * Share packaging and release management infrastructure >> > >> > * Reduce / eliminate problems due to API changes (where we currently >> > introduce breakage into our CI workflow when there is a breaking / >> > incompatible change) >> > >> > * Arrow releases would include a coordinated snapshot of the Parquet >> > implementation as it stands >> > >> > Continuing with the status quo has become unsatisfactory to me and as >> > a result I've become less motivated to work on the parquet-cpp >> > codebase. >> > >> > The only Parquet C++ committer who is not an Arrow committer is Deepak >> > Majeti. I think the issue of commit privileges could be resolved >> > without too much difficulty or time. >> > >> > I also think if it is truly necessary that the Apache Parquet >> > community could create release scripts to cut a miniml versioned >> > Apache Parquet C++ release if
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
hi Deepak, responses inline On Sun, Jul 29, 2018 at 10:44 PM, Deepak Majeti wrote: > I dislike the current build system complications as well. > > However, in my opinion, combining the code bases will severely impact the > progress of the parquet-cpp project and implicitly the progress of the > entire parquet project. > Combining would have made much more sense if parquet-cpp is a mature > project and codebase. But parquet-cpp (and the entire parquet project) is > evolving continuously with new features being added including bloom > filters, column encryption, and indexes. > I don't see why parquet-cpp development would be impacted in a negative way. In fact, I've argued exactly the opposite. Can you explain in more detail why you think this would be the case? If anything, parquet-cpp would benefit from more mature and better maintained developer infrastructure. Here's the project shortlog: $ git shortlog -sn 08acdf6bfe3cd160ffe19b79bbded2bdc3f7bd62..master 145 Wes McKinney 109 Uwe L. Korn 53 Deepak Majeti 38 Korn, Uwe 36 Nong Li 12 Kouhei Sutou 10 Max Risuhin 9 Antoine Pitrou 8 rip.nsk 6 Phillip Cloud 6 Xianjin YE 5 Aliaksei Sandryhaila 4 Thomas Sanchez 3 Artem Tarasov 3 Joshua Storck 3 Lars Volker 3 fscheibner 3 revaliu 2 Itai Incze 2 Kalon Mills 2 Marc Vertes 2 Mike Trinkala 2 Philipp Hoch 1 Alec Posney 1 Christopher C. Aycock 1 Colin Nichols 1 Dmitry Bushev 1 Eric Daniel 1 Fabrizio Fabbri 1 Florian Scheibner 1 Jaguar Xiong 1 Julien Lafaye 1 Julius Neuffer 1 Kashif Rasul 1 Rene Sugar 1 Robert Gruener 1 Toby Shaw 1 William Forson 1 Yue Chen 1 thamht4190 Out of these, I know for a fact that at least the following contributed to parquet-cpp as a result of their involvement with Apache Arrow: 145 Wes McKinney 109 Uwe L. Korn 38 Korn, Uwe 12 Kouhei Sutou 10 Max Risuhin 9 Antoine Pitrou 6 Phillip Cloud 3 Joshua Storck 1 Christopher C. Aycock 1 Rene Sugar 1 Robert Gruener This is ~70% of commits > If the two code bases merged, it will be much more difficult to contribute > to the parquet-cpp project since now Arrow bindings have to be supported as > well. Please correct me if I am wrong here. I don't see why this would be true. The people above are already supporting these bindings (which are pretty isolated to the symbols in the parquet::arrow namespace), and patches not having to do with the Arrow columnar data structures would not be affected. Because of the arguments I made in my first e-mail, it will be less work for the developers working on both projects to maintain the interfaces. Currently, it is necessary to make patches to multiple projects to improve APIs and fix bugs in many cases. > > Out of the two evils, I think handling the build system, packaging > duplication is much more manageable since they are quite stable at this > point. We've been talking about this for a long time and no concrete and actionable solution has come forward. > > Regarding "* API changes cause awkward release coordination issues between > Arrow and Parquet". Can we make minor releases for parquet-cpp (with API > changes needed) as and when Arrow is released? The central issue is that changes frequently require changes to multiple codebases and cross-repo CI to verify patches jointly is not really possible > > Regarding "we maintain a Arrow conversion code in parquet-cpp for > converting between Arrow columnar memory format and Parquet". Can this be > moved to the Arrow project and expose the more stable low-level APIs in > parquet-cpp? The parts of Parquet that do not interact with the Arrow columnar format still use Arrow platform APIs (IO, memory management, compression, algorithms, etc.). We still would therefore have a circular dependency, though some parts (e.g. changes in the parquet::arrow layer) might be easier - Wes > > I am also curious if the Arrow and Parquet Java implementations have > similar API compatibility issues. Parquet-Java is pretty different: * On the plus side, the built-in Java platform solves some of the problems we have addressed in the Arrow platform APIs. Note that Arrow hasn't reinvented any wheels here or failed to use tools available in the C++ standard library or Boost -- if you look at major Google codebases like TensorFlow they have developed nearly identical platform APIs to solve the same problems * Parquet-Java depends on Hadoop platform APIs which has caused problems for other Java projects which wish to read and write Parquet files but do not use Hadoop (e.g. they store data in S3) > > > On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney wrote: > >> hi folks, >> >> We've been struggling for quite some time with the development >> workflow between the Arrow and Parquet C++ (and
Re: [DISCUSS] Solutions for improving the Arrow-Parquet C++ development morass
I dislike the current build system complications as well. However, in my opinion, combining the code bases will severely impact the progress of the parquet-cpp project and implicitly the progress of the entire parquet project. Combining would have made much more sense if parquet-cpp is a mature project and codebase. But parquet-cpp (and the entire parquet project) is evolving continuously with new features being added including bloom filters, column encryption, and indexes. If the two code bases merged, it will be much more difficult to contribute to the parquet-cpp project since now Arrow bindings have to be supported as well. Please correct me if I am wrong here. Out of the two evils, I think handling the build system, packaging duplication is much more manageable since they are quite stable at this point. Regarding "* API changes cause awkward release coordination issues between Arrow and Parquet". Can we make minor releases for parquet-cpp (with API changes needed) as and when Arrow is released? Regarding "we maintain a Arrow conversion code in parquet-cpp for converting between Arrow columnar memory format and Parquet". Can this be moved to the Arrow project and expose the more stable low-level APIs in parquet-cpp? I am also curious if the Arrow and Parquet Java implementations have similar API compatibility issues. On Sat, Jul 28, 2018 at 7:45 PM Wes McKinney wrote: > hi folks, > > We've been struggling for quite some time with the development > workflow between the Arrow and Parquet C++ (and Python) codebases. > > To explain the root issues: > > * parquet-cpp depends on "platform code" in Apache Arrow; this > includes file interfaces, memory management, miscellaneous algorithms > (e.g. dictionary encoding), etc. Note that before this "platform" > dependency was introduced, there was significant duplicated code > between these codebases and incompatible abstract interfaces for > things like files > > * we maintain a Arrow conversion code in parquet-cpp for converting > between Arrow columnar memory format and Parquet > > * we maintain Python bindings for parquet-cpp + Arrow interop in > Apache Arrow. This introduces a circular dependency into our CI. > > * Substantial portions of our CMake build system and related tooling > are duplicated between the Arrow and Parquet repos > > * API changes cause awkward release coordination issues between Arrow > and Parquet > > I believe the best way to remedy the situation is to adopt a > "Community over Code" approach and find a way for the Parquet and > Arrow C++ development communities to operate out of the same code > repository, i.e. the apache/arrow git repository. > > This would bring major benefits: > > * Shared CMake build infrastructure, developer tools, and CI > infrastructure (Parquet is already being built as a dependency in > Arrow's CI systems) > > * Share packaging and release management infrastructure > > * Reduce / eliminate problems due to API changes (where we currently > introduce breakage into our CI workflow when there is a breaking / > incompatible change) > > * Arrow releases would include a coordinated snapshot of the Parquet > implementation as it stands > > Continuing with the status quo has become unsatisfactory to me and as > a result I've become less motivated to work on the parquet-cpp > codebase. > > The only Parquet C++ committer who is not an Arrow committer is Deepak > Majeti. I think the issue of commit privileges could be resolved > without too much difficulty or time. > > I also think if it is truly necessary that the Apache Parquet > community could create release scripts to cut a miniml versioned > Apache Parquet C++ release if that is deemed truly necessary. > > I know that some people are wary of monorepos and megaprojects, but as > an example TensorFlow is at least 10 times as large of a projects in > terms of LOCs and number of different platform components, and it > seems to be getting along just fine. I think we should be able to work > together as a community to function just as well. > > Interested in the opinions of others, and any other ideas for > practical solutions to the above problems. > > Thanks, > Wes > -- regards, Deepak Majeti