[Factor-talk] Quick Question For Chris Double
Chris -- I'm spending some time going back through various key bloggers' contributions and am working through some of yours at the moment. Back in December of 2006 in one of your articles on Parser Combinators you reference a Chapter 5 on Parser Combinators from some larger work. This is a nice article and it leads to wonder what the source is that it is taken from and whether the surrounding chapters are any good and worth reading as well. Your thoughts? Mike-- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Quick Question For Chris Double
On Fri, Aug 24, 2012 at 11:29 PM, Michael Clagett mclag...@hotmail.com wrote: Back in December of 2006 in one of your articles on Parser Combinators you reference a Chapter 5 on Parser Combinators from some larger work. This is a nice article and it leads to wonder what the source is that it is taken from and whether the surrounding chapters are any good and worth reading as well. I that that was this chapter: http://www.bluishcoder.co.nz/II.05.ParserCombinators.pdf It was from Part II of a book on the Clean programming language. Unfortunately it seems to no longer be available according to: http://wiki.clean.cs.ru.nl/Functional_Programming_in_Clean Luckily it's rare for things to disappear completely from the net: http://www.mmnt.net/db/0/0/ftp.cs.kun.nl/pub/CSI/SoftwEng.FunctLang/papers/cleanbook The case studies are a good read. Chris. -- http://www.bluishcoder.co.nz -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Quick Question For Chris Double
Thanks much. In this age of erroneous credit reports, that never disappearing aspect of the web is definitely a double-edged sword. Sent from my iPhone On Aug 24, 2012, at 9:41 AM, Chris Double chris.dou...@double.co.nz wrote: On Fri, Aug 24, 2012 at 11:29 PM, Michael Clagett mclag...@hotmail.com wrote: Back in December of 2006 in one of your articles on Parser Combinators you reference a Chapter 5 on Parser Combinators from some larger work. This is a nice article and it leads to wonder what the source is that it is taken from and whether the surrounding chapters are any good and worth reading as well. I that that was this chapter: http://www.bluishcoder.co.nz/II.05.ParserCombinators.pdf It was from Part II of a book on the Clean programming language. Unfortunately it seems to no longer be available according to: http://wiki.clean.cs.ru.nl/Functional_Programming_in_Clean Luckily it's rare for things to disappear completely from the net: http://www.mmnt.net/db/0/0/ftp.cs.kun.nl/pub/CSI/SoftwEng.FunctLang/papers/cleanbook The case studies are a good read. Chris. -- http://www.bluishcoder.co.nz -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
On Fri, Aug 24, 2012 at 12:33 AM, Tim Allen screwt...@froup.com wrote: I'm no expert in these matters, but it worries me a little that you mention Ruby Gems as one potential inspiration: I know that (for example) Debian doesn't package many Ruby libraries because the Gem system interacts poorly with a system-wide package-manager. Other languages like Perl and Python seem to have tons of libraries packaged in Debian, so they might be better models. I'd call this a plus, because Debian's update schedule is glacial, and you don't want to be limited by your system package manager when installing a library for yourself for a local project. Systems like gem and npm let you easily install up-to-date packages at a user- or project-local location without requiring admin access or coordination with the system's update schedule. Having a image.factorimage file in a package might make life a bit complicated - aren't Factor images platform specific? What happens if somebody launches the Factor VM from inside the package directory but the image is for a different architecture? Images would not be distributed with packages. `image.factorimage` would just be a standard default location for `save-image` and image loading during package development. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
Python's package management is (typically) on a system basis while afaik npm and gem is per project and user respectively, so this seems to be comparing apples and oranges. For example, there is no standard way to download a Python project and do the equivalent of npm install. Some Linux distros do manage to keep their python packages current, but the Python project I work on (with 30+ dependencies) distributes all the deps with the package or requests users to use pip or easy_install because version numbers of Python packages across Linuxes seem to follow a log normal distribution. On Fri, Aug 24, 2012 at 8:14 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 12:33 AM, Tim Allen screwt...@froup.com wrote: I'm no expert in these matters, but it worries me a little that you mention Ruby Gems as one potential inspiration: I know that (for example) Debian doesn't package many Ruby libraries because the Gem system interacts poorly with a system-wide package-manager. Other languages like Perl and Python seem to have tons of libraries packaged in Debian, so they might be better models. I'd call this a plus, because Debian's update schedule is glacial, and you don't want to be limited by your system package manager when installing a library for yourself for a local project. Systems like gem and npm let you easily install up-to-date packages at a user- or project-local location without requiring admin access or coordination with the system's update schedule. Having a image.factorimage file in a package might make life a bit complicated - aren't Factor images platform specific? What happens if somebody launches the Factor VM from inside the package directory but the image is for a different architecture? Images would not be distributed with packages. `image.factorimage` would just be a standard default location for `save-image` and image loading during package development. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
Python's package management is terrible. They have easy_install, distribute, distribute2, pip, and virtualenv, in addition to having binaries/interpreters/libraries for python2 and python3. I tried pip but it couldn't install one of the packages I needed, so I wiped them all and went back to easy_install. Everything is global and I find myself having to edit config files and rm directories/egg files by hand to upgrade packages. I had a .11-git release of a library, and even though .11 was out, easy_install wouldn't upgrade it. You have to micromanage each installed package. If you want a model of what not to do, look to python. Doug On Fri, Aug 24, 2012 at 11:35 AM, Marmaduke Woodman mmwood...@gmail.com wrote: Python's package management is (typically) on a system basis while afaik npm and gem is per project and user respectively, so this seems to be comparing apples and oranges. For example, there is no standard way to download a Python project and do the equivalent of npm install. Some Linux distros do manage to keep their python packages current, but the Python project I work on (with 30+ dependencies) distributes all the deps with the package or requests users to use pip or easy_install because version numbers of Python packages across Linuxes seem to follow a log normal distribution. On Fri, Aug 24, 2012 at 8:14 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 12:33 AM, Tim Allen screwt...@froup.com wrote: I'm no expert in these matters, but it worries me a little that you mention Ruby Gems as one potential inspiration: I know that (for example) Debian doesn't package many Ruby libraries because the Gem system interacts poorly with a system-wide package-manager. Other languages like Perl and Python seem to have tons of libraries packaged in Debian, so they might be better models. I'd call this a plus, because Debian's update schedule is glacial, and you don't want to be limited by your system package manager when installing a library for yourself for a local project. Systems like gem and npm let you easily install up-to-date packages at a user- or project-local location without requiring admin access or coordination with the system's update schedule. Having a image.factorimage file in a package might make life a bit complicated - aren't Factor images platform specific? What happens if somebody launches the Factor VM from inside the package directory but the image is for a different architecture? Images would not be distributed with packages. `image.factorimage` would just be a standard default location for `save-image` and image loading during package development. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
Here's an unpopular vote, but I'll say it anyway. How about portable packages? As in, if you download package P that is dependent on packages D, E, F, then package P comes with the correct versions of D, E, F embedded in it? That way we do away with dependencies altogether. (yes, there'll be redundancy for the sake of convenience. I don't know the implications re compiling) And we can pass the bucket to the programmer when P1's E (version 1.1) returns an object that needs to be passed to P2's E (version 1.2) - now the incompatibility issue is on the surface, and the programmer can look at both sides, see where they disagree, and code up a bridge. Or better yet, E can deprecate its modified functions by appending the version number to their name. Either way, the main idea is: do away with dependencies, make packages portable. (or at least give them an interface between the package and its dependencies, like what I'm told OCaml does - or was it SML? or both?) To me the biggest problem with package systems is dependencies. I don't know why you folks are discussing Linux to be honest. - rien On Fri, Aug 24, 2012 at 2:46 PM, Doug Coleman doug.cole...@gmail.comwrote: Python's package management is terrible. They have easy_install, distribute, distribute2, pip, and virtualenv, in addition to having binaries/interpreters/libraries for python2 and python3. I tried pip but it couldn't install one of the packages I needed, so I wiped them all and went back to easy_install. Everything is global and I find myself having to edit config files and rm directories/egg files by hand to upgrade packages. I had a .11-git release of a library, and even though .11 was out, easy_install wouldn't upgrade it. You have to micromanage each installed package. If you want a model of what not to do, look to python. Doug On Fri, Aug 24, 2012 at 11:35 AM, Marmaduke Woodman mmwood...@gmail.com wrote: Python's package management is (typically) on a system basis while afaik npm and gem is per project and user respectively, so this seems to be comparing apples and oranges. For example, there is no standard way to download a Python project and do the equivalent of npm install. Some Linux distros do manage to keep their python packages current, but the Python project I work on (with 30+ dependencies) distributes all the deps with the package or requests users to use pip or easy_install because version numbers of Python packages across Linuxes seem to follow a log normal distribution. On Fri, Aug 24, 2012 at 8:14 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 12:33 AM, Tim Allen screwt...@froup.com wrote: I'm no expert in these matters, but it worries me a little that you mention Ruby Gems as one potential inspiration: I know that (for example) Debian doesn't package many Ruby libraries because the Gem system interacts poorly with a system-wide package-manager. Other languages like Perl and Python seem to have tons of libraries packaged in Debian, so they might be better models. I'd call this a plus, because Debian's update schedule is glacial, and you don't want to be limited by your system package manager when installing a library for yourself for a local project. Systems like gem and npm let you easily install up-to-date packages at a user- or project-local location without requiring admin access or coordination with the system's update schedule. Having a image.factorimage file in a package might make life a bit complicated - aren't Factor images platform specific? What happens if somebody launches the Factor VM from inside the package directory but the image is for a different architecture? Images would not be distributed with packages. `image.factorimage` would just be a standard default location for `save-image` and image loading during package development. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list
Re: [Factor-talk] Modules/packages for Factor
rien, Read this blog post: http://www.mikealrogers.com/posts/nodemodules-in-git.html He makes a very convincing argument against bundling packages. Doug On Fri, Aug 24, 2012 at 11:56 AM, P. uploa...@gmail.com wrote: Here's an unpopular vote, but I'll say it anyway. How about portable packages? As in, if you download package P that is dependent on packages D, E, F, then package P comes with the correct versions of D, E, F embedded in it? That way we do away with dependencies altogether. (yes, there'll be redundancy for the sake of convenience. I don't know the implications re compiling) And we can pass the bucket to the programmer when P1's E (version 1.1) returns an object that needs to be passed to P2's E (version 1.2) - now the incompatibility issue is on the surface, and the programmer can look at both sides, see where they disagree, and code up a bridge. Or better yet, E can deprecate its modified functions by appending the version number to their name. Either way, the main idea is: do away with dependencies, make packages portable. (or at least give them an interface between the package and its dependencies, like what I'm told OCaml does - or was it SML? or both?) To me the biggest problem with package systems is dependencies. I don't know why you folks are discussing Linux to be honest. - rien On Fri, Aug 24, 2012 at 2:46 PM, Doug Coleman doug.cole...@gmail.com wrote: Python's package management is terrible. They have easy_install, distribute, distribute2, pip, and virtualenv, in addition to having binaries/interpreters/libraries for python2 and python3. I tried pip but it couldn't install one of the packages I needed, so I wiped them all and went back to easy_install. Everything is global and I find myself having to edit config files and rm directories/egg files by hand to upgrade packages. I had a .11-git release of a library, and even though .11 was out, easy_install wouldn't upgrade it. You have to micromanage each installed package. If you want a model of what not to do, look to python. Doug On Fri, Aug 24, 2012 at 11:35 AM, Marmaduke Woodman mmwood...@gmail.com wrote: Python's package management is (typically) on a system basis while afaik npm and gem is per project and user respectively, so this seems to be comparing apples and oranges. For example, there is no standard way to download a Python project and do the equivalent of npm install. Some Linux distros do manage to keep their python packages current, but the Python project I work on (with 30+ dependencies) distributes all the deps with the package or requests users to use pip or easy_install because version numbers of Python packages across Linuxes seem to follow a log normal distribution. On Fri, Aug 24, 2012 at 8:14 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 12:33 AM, Tim Allen screwt...@froup.com wrote: I'm no expert in these matters, but it worries me a little that you mention Ruby Gems as one potential inspiration: I know that (for example) Debian doesn't package many Ruby libraries because the Gem system interacts poorly with a system-wide package-manager. Other languages like Perl and Python seem to have tons of libraries packaged in Debian, so they might be better models. I'd call this a plus, because Debian's update schedule is glacial, and you don't want to be limited by your system package manager when installing a library for yourself for a local project. Systems like gem and npm let you easily install up-to-date packages at a user- or project-local location without requiring admin access or coordination with the system's update schedule. Having a image.factorimage file in a package might make life a bit complicated - aren't Factor images platform specific? What happens if somebody launches the Factor VM from inside the package directory but the image is for a different architecture? Images would not be distributed with packages. `image.factorimage` would just be a standard default location for `save-image` and image loading during package development. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can
Re: [Factor-talk] Modules/packages for Factor
On Fri, Aug 24, 2012 at 2:56 PM, P. uploa...@gmail.com wrote: Here's an unpopular vote, but I'll say it anyway. How about portable packages? As in, if you download package P that is dependent on packages D, E, F, then package P comes with the correct versions of D, E, F embedded in it? That way we do away with dependencies altogether. (yes, there'll be redundancy for the sake of convenience. I don't know the implications re compiling) As Doug noted, generally you don't want to bundle dependencies. However, I believe the proposed would handle this use case if you really wanted to, since you could distribute a populated `packages` subdirectory inside your top-level package's distribution. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? On Fri, Aug 24, 2012 at 3:02 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 2:56 PM, P. uploa...@gmail.com wrote: Here's an unpopular vote, but I'll say it anyway. How about portable packages? As in, if you download package P that is dependent on packages D, E, F, then package P comes with the correct versions of D, E, F embedded in it? That way we do away with dependencies altogether. (yes, there'll be redundancy for the sake of convenience. I don't know the implications re compiling) As Doug noted, generally you don't want to bundle dependencies. However, I believe the proposed would handle this use case if you really wanted to, since you could distribute a populated `packages` subdirectory inside your top-level package's distribution. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
Joe, I was looking for something more enforced than that - so that I can expect a package I download to just work and not waste time with dependence issues. Doug, I just read that link but it's a bit over my head for me, not knowing node, and not having any concrete examples of how it's a bad thing (like the example I gave with P and E, D, F). So it's hard for me to follow. - rien On Fri, Aug 24, 2012 at 3:05 PM, P. uploa...@gmail.com wrote: Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? On Fri, Aug 24, 2012 at 3:02 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 2:56 PM, P. uploa...@gmail.com wrote: Here's an unpopular vote, but I'll say it anyway. How about portable packages? As in, if you download package P that is dependent on packages D, E, F, then package P comes with the correct versions of D, E, F embedded in it? That way we do away with dependencies altogether. (yes, there'll be redundancy for the sake of convenience. I don't know the implications re compiling) As Doug noted, generally you don't want to bundle dependencies. However, I believe the proposed would handle this use case if you really wanted to, since you could distribute a populated `packages` subdirectory inside your top-level package's distribution. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
On Fri, Aug 24, 2012 at 3:05 PM, P. uploa...@gmail.com wrote: Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? An idea I had was to use test suites to describe dependencies. You could say if this package implementation passes this test suite, then it fulfills this dependency. Whether it's practical to write detailed enough test suites in practice is a good question, but in theory it could allow you to express implementation requirements in a more flexible way than module signatures or version numbers. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
On Fri, Aug 24, 2012 at 3:13 PM, P. uploa...@gmail.com wrote: Joe, I was looking for something more enforced than that - so that I can expect a package I download to just work and not waste time with dependence issues. A package metadata file could perhaps have a `self-contained` flag that would restrict the search path to only the package-local and standard library directories. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
I remember your describing that idea! I think it's great - it would allow us to express the interface not only in a more flexible way as you put it, but in a more detailed way (much more than a type system could provide). You have my vote. What are the downsides? (there's no language doing that now, right?) - rien On Fri, Aug 24, 2012 at 3:13 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 3:05 PM, P. uploa...@gmail.com wrote: Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? An idea I had was to use test suites to describe dependencies. You could say if this package implementation passes this test suite, then it fulfills this dependency. Whether it's practical to write detailed enough test suites in practice is a good question, but in theory it could allow you to express implementation requirements in a more flexible way than module signatures or version numbers. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
We have a code coverage tool. If we started using it, one could at least be guaranteed that every code path is at least hit by the unit tests. Doug On Fri, Aug 24, 2012 at 12:27 PM, P. uploa...@gmail.com wrote: I remember your describing that idea! I think it's great - it would allow us to express the interface not only in a more flexible way as you put it, but in a more detailed way (much more than a type system could provide). You have my vote. What are the downsides? (there's no language doing that now, right?) - rien On Fri, Aug 24, 2012 at 3:13 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 3:05 PM, P. uploa...@gmail.com wrote: Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? An idea I had was to use test suites to describe dependencies. You could say if this package implementation passes this test suite, then it fulfills this dependency. Whether it's practical to write detailed enough test suites in practice is a good question, but in theory it could allow you to express implementation requirements in a more flexible way than module signatures or version numbers. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Modules/packages for Factor
I think there is a danger in proliferation of too many packages with too many versions here. One advantage that Factor has had, albeit a mixed advantage due to reduced resources at times, is one main codebase that can be improved in wide cross-cutting ways. That has resulted in a lot of common vocabulary improvements instead of forks. We should support packages, and versions, and freezing of requirements for your application, but lets not lose sight of the benefits we've enjoyed thus far. As an aside, the CPAN testing infrastructure is something to be copied. Every time someone uploads a new module it goes through various automated testing on different architectures, and produces a build report back to you. That combined with some kind of cross-module lint tool might be able to yield the strong commons that I appreciate in Factor. Best, John. P.S. Let's not paint ourselves into a dependency hell situation either... On Fri, Aug 24, 2012 at 12:27 PM, P. uploa...@gmail.com wrote: I remember your describing that idea! I think it's great - it would allow us to express the interface not only in a more flexible way as you put it, but in a more detailed way (much more than a type system could provide). You have my vote. What are the downsides? (there's no language doing that now, right?) - rien On Fri, Aug 24, 2012 at 3:13 PM, Joe Groff arc...@gmail.com wrote: On Fri, Aug 24, 2012 at 3:05 PM, P. uploa...@gmail.com wrote: Hey Doug, thanks for the link, I'll read it when I'm not at work. (oops :P) What about mimicking something like OCaml's functors or whatever they call their dependency interfaces? An idea I had was to use test suites to describe dependencies. You could say if this package implementation passes this test suite, then it fulfills this dependency. Whether it's practical to write detailed enough test suites in practice is a good question, but in theory it could allow you to express implementation requirements in a more flexible way than module signatures or version numbers. -Joe -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
[Factor-talk] Out of memory error
Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Out of memory error
How much ram? On Aug 24, 2012 6:01 PM, Alexander J. Vondrak ajvond...@csupomona.edu wrote: Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Out of memory error
8 GB. From: Doug Coleman [doug.cole...@gmail.com] Sent: Friday, August 24, 2012 6:06 PM To: factor-talk@lists.sourceforge.net Subject: Re: [Factor-talk] Out of memory error How much ram? On Aug 24, 2012 6:01 PM, Alexander J. Vondrak ajvond...@csupomona.edumailto:ajvond...@csupomona.edu wrote: Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.netmailto:Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Out of memory error
8gb outta be enough for anybody. But It's not, I guess. Try with more RAM or patch the part that grows to only grow to as much as you have. Maybe something is using way too much On Aug 24, 2012 6:01 PM, Alexander J. Vondrak ajvond...@csupomona.edu wrote: Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Out of memory error
It seems to fail soon after the tools.deploy tests. I tried at one point disabling those, but I think it still crashed. I should check again... From: Doug Coleman [doug.cole...@gmail.com] Sent: Friday, August 24, 2012 6:44 PM To: factor-talk@lists.sourceforge.net Subject: Re: [Factor-talk] Out of memory error 8gb outta be enough for anybody. But It's not, I guess. Try with more RAM or patch the part that grows to only grow to as much as you have. Maybe something is using way too much On Aug 24, 2012 6:01 PM, Alexander J. Vondrak ajvond...@csupomona.edumailto:ajvond...@csupomona.edu wrote: Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.netmailto:Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk
Re: [Factor-talk] Out of memory error
It might be worth only loading and testing vocabularies in core and basis, to see if GVN works? On Aug 24, 2012, at 6:50 PM, Alexander J. Vondrak ajvond...@csupomona.edu wrote: It seems to fail soon after the tools.deploy tests. I tried at one point disabling those, but I think it still crashed. I should check again... From: Doug Coleman [doug.cole...@gmail.com] Sent: Friday, August 24, 2012 6:44 PM To: factor-talk@lists.sourceforge.net Subject: Re: [Factor-talk] Out of memory error 8gb outta be enough for anybody. But It's not, I guess. Try with more RAM or patch the part that grows to only grow to as much as you have. Maybe something is using way too much On Aug 24, 2012 6:01 PM, Alexander J. Vondrak ajvond...@csupomona.edumailto:ajvond...@csupomona.edu wrote: Hey all, Mini-update: got the unit tests for the global value numbering pass more up-to-snuff awhile ago (it's like test-driven development in reverse!). Still seems like I should be doing more to test the new capabilities, but I'm not sure if I could do much more than what's there. My efforts as of late have been focused on some GVN improvements. But for the past way-too-long, I thought I had broken something catastrophic because `load-all test-all` would invariably crash the listener with an out of memory error: Out of memory Nursery: Start=831c, size=10, end=832c Aging: Start=82fc, size=20, end=831c Tenured: Start=50c8, size=3214, end=82dc Cards:base=b44d0008, size=326800 Aborted After a lot of very slow debugging (them's the breaks when running all the unit tests, I guess) that was ultimately fruitless, I finally tried out Factor 0.95 straight from the build farm---none of my changes, no .git, just a clean download of the linux-x86-32 tarchive. And, what do you know, `load-all test-all` still crashes with the same error. It hadn't done that before, so I think the issue may be an intervening OS update I did. I'm running a fully up-to-date Debian testing, kernel 3.2.0-3-686-pae. 1) Can anyone reproduce this error? Or is my machine just that jacked up? I'd be happy to provide other diagnostic info, if it'd help. 2) Any coping mechanisms? I considered just throwing a bunch of memory at it with command line switches (-young=, -aging=, -tenured=, etc.), but wasn't really sure how much would do the trick, since it just started crashing like this out of the blue anyway. Thanks, --Alex Vondrak -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.netmailto:Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk -- Live Security Virtual Conference Exclusive live event will cover all the ways today's security and threat landscape has changed and how IT managers can respond. Discussions will include endpoint security, mobile security and the latest in malware threats. http://www.accelacomm.com/jaw/sfrnl04242012/114/50122263/ ___ Factor-talk mailing list Factor-talk@lists.sourceforge.net https://lists.sourceforge.net/lists/listinfo/factor-talk