Re: using git directory cache code in darcs?
> Not until all the data structures are really really stable. Fine by me to wait, though perhaps not for the same reason, and perhaps not as long. A libgit.so can deal with data structure changes just as well as a set of command line utilities. So long as everything funnels through one place, you can change by changing that one place. However I am willing to agree that its not libgit time yet, for two reasons: 1) everyone who has two clues on the subject is too busy and too productive on more pressing git issues, and 2) in addition to internal data structures being not yet stable, I suspect that the operations (git commands, options and behaviour) are also not stable. The first step of a good libgit is not coding to the internal data structures, but rather designing the interface (the operations, arguments, data types, and behaviour). So, until people have time, and the interface ops are settled down, its too early to design libgit. Or at least too early to publish a design and seek concensus. If I had the time the first thing I'd be doing right now would be designing libgit on the side, anticipating the day when it was time to publish a draft and engage the community discussion that leads to an adequate concensus. === By the way, a good libgit design, in my view, would isolate the data structures written to files below .git from the data structures presented at the library API, to some extent. Changes in the file structures must be handled without disrupting the library API. If a libgit API didn't isolate the library caller from details of the structures in files below .git, then yes you'd want really really stable data structures, impossibly stable in fact. That way leads to hacks and workarounds in the future, because the data structures are never perfectly stable. -- I won't rest till it's the best ... Programmer, Linux Scalability Paul Jackson <[EMAIL PROTECTED]> 1.650.933.1373, 1.925.600.0401 - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
On Sun, 17 Apr 2005, Paul Jackson wrote: > > I'd encourage you to invite someone to provide a libgit. Not until all the data structures are really really stable. That's the thing - we can keep the _program_ interfaces somewhat stable. But internally we may change stuff wildly, and anybody who depends on a library interface would be screwed. Ergo: no library interfaces yet. Wait for it to stabilize. Start trying to just script the programs. Linus - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
Linus wrote: > But then I'd really suggest that you use "git" itself, not any > "libgit". Ie you take _all_ the plumbing as real programs, and instead of > trying to link against individual routines, you'd _script_ it. I think you've got this upside down, Linus. Trying to make the executable 'git' commands the bottom layer of the user implementation stack forces inefficiencies on higher layers of the stack, and thus encourages stupid workarounds and cheats in an effort to speed things up. I'd encourage you to invite someone to provide a libgit. Such work should _start_ with proposing and gaining acceptance on the API - the calls, the arguments, the types, the rough idea of the semantics. The actual coding is the easy part. One is not slave to the agreed API when coding. The API will continue to evolve, but if the originally accepted proposal was sound, the evolution will be at a modest rate, with few incompatibilities introduced. If several operations should be done as a unit, to preserve the integrity of the .git data or to provide sane results, then libgit need only provide such pre-packaged units, not the incomplete fragments from which they are composed. That is, the libgit calls could quite possibly be at roughly the same semantic level as your git commands. One could even code up some of the libgit calls, in early versions of libgit, by simply invoking the corresponding git command. But, eventually, all the git commands should be recoded on top of the libgit library, and the libgit library become the canonical user interface to git, on which all else is layered. One typical way that this choice manifests itself is in the strace output from doing some ordinary git command from a C program that is implementing an SCM system on top of git. Forcing every operation to be done via a separate git command execution mushrooms the number of kernel system calls a hundred fold, or two hundred fold if some dang fool uses system(3S) to invoke the git command. What might have been a handful of calls to stat/open/read/write/close a file turns into a mini-shell session. That way lies insanity, or at least painful inefficiency, and the usual parade of bugs, stupid coding tricks and painful user interfaces that follow in the wake. The recommended layering of such user facilities is well known, with a C library at the bottom. Granted, the history of source code management tools provides few examples of this recommended layering. On top of this library go plugin modules for the fancier scripting languages that accept such. Swig can be used to aid this construction, for Tcl, Python, Perl, Guile, Java, Ruby, Mzscheme, PHP, Ocaml, Pike, C#, Chicken Allegro CL, Modula-3, Javascript and Eiffel. Though I personally have not worked with Swig enough to gain success with it. The only such modules I've done were handcoded Python modules. Also on top of this library one provides a set of command line utilities or one multiplexed 'git foo ...' command, for use at shell prompts. Or the command line utilities can be coded in one of the above higher level scripting languages, using in turn the git library plugin. However many of these scripting languages bring runtime requirements that are not universally satisfied on all target systems, so are a poor choice for this purpose. If I am recalling correctly, from the days when I regularly used bk, one of the things that Larry did right with bk, which RCS and SCCS did not do right before then, was to provide a low level library to his storage - a cleanroom recoded variant of SCCS in his case. Implementing production source control systems on top of a set of executable commands is a pain in the arse. An all too familiar pain. I'd repeat my encouragement that you invite someone to provide such a libgit, however since I have other commitments for the next month at least, so can't volunteer right away, if ever, it is more appropriate that I shut up now, under the old "put up code or be quiet" rule. -- I won't rest till it's the best ... Programmer, Linux Scalability Paul Jackson <[EMAIL PROTECTED]> 1.650.933.1373, 1.925.600.0401 - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
Junio wrote: > What Linus is saying is that he wants you to talk with git > plumbing by invoking the executables he have, via system(3), > popen(3), etc. Hopefully, Linus didn't specify system(3) or popen(3) for production software. They are a rich source of security holes. Inefficient, too, since they invoke a shell process to interpret the command. Use execve(2), or exevl(3), execle(3), execv(3). Or if you really enjoy the path search, use execlp or execvp, but with your own $PATH, not trusting the one passed in via the environment any further than you can throw it. However, on further consideration, I think Linus is wrong to recommend that the git executables, not a libgit library, be the 'basic user level on which all else is based." I will reply to a Linus post, expounding on that thought further. -- I won't rest till it's the best ... Programmer, Linux Scalability Paul Jackson <[EMAIL PROTECTED]> 1.650.933.1373, 1.925.600.0401 - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
> "NA" == Nomad Arton <[EMAIL PROTECTED]> writes: NA> Linus Torvalds schrieb: >> In fact, one of my hopes was that other SCM's could just use the git >> plumbing. But then I'd really suggest that you use "git" itself, not any >> "libgit". Ie you take _all_ the plumbing as real programs, and instead of >> trying to link against individual routines, you'd _script_ it. NA> please excuse; libgit and scripting to me arent a contradiction. many NA> sripting languages are extended by C modules, while still happening to NA> have all the scripting rapidity. its just a matter of how to NA> communicate with the C code, isnt it? You are arguing for scripting language binding like what Swig creates. While that would also be a worthy addition, having language binding is not the only way to do _script_. What Linus is saying is that he wants you to talk with git plumbing by invoking the executables he have, via system(3), popen(3), etc. The C-level first has to be libified before you can start talking about host language bindings but that just started to happen and is not ready yet. However, you can use and benefit from GIT without waiting for that kind of integration, if you use the "spawning the executables" approach. I agree with him. - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
Linus Torvalds schrieb: In fact, one of my hopes was that other SCM's could just use the git plumbing. But then I'd really suggest that you use "git" itself, not any "libgit". Ie you take _all_ the plumbing as real programs, and instead of trying to link against individual routines, you'd _script_ it. please excuse; libgit and scripting to me arent a contradiction. many sripting languages are extended by C modules, while still happening to have all the scripting rapidity. its just a matter of how to communicate with the C code, isnt it? yours, peter - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
Linus Torvalds wrote: On Sun, 17 Apr 2005, David Roundy wrote: That's all right. Darcs would only access the cached data through a git-caching layer, and we've already got an abstraction layer over the pristine cache. As long as the git layer can quickly retrieve the contents of a given file, we should be fine. Yes. In fact, one of my hopes was that other SCM's could just use the git plumbing. But then I'd really suggest that you use "git" itself, not any "libgit". Ie you take _all_ the plumbing as real programs, and instead of trying to link against individual routines, you'd _script_ it. If you don't want it, I won't do it. Still makes sense to separate the plumbing from the porcelain, though. -- Mike Taht "You can tell how far we have to go, when FORTRAN is the language of supercomputers. -- Steven Feiner" - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
On Sun, 17 Apr 2005, David Roundy wrote: > > That's all right. Darcs would only access the cached data through a > git-caching layer, and we've already got an abstraction layer over the > pristine cache. As long as the git layer can quickly retrieve the contents > of a given file, we should be fine. Yes. In fact, one of my hopes was that other SCM's could just use the git plumbing. But then I'd really suggest that you use "git" itself, not any "libgit". Ie you take _all_ the plumbing as real programs, and instead of trying to link against individual routines, you'd _script_ it. In other words, "git" would be an independent cache of the real SCM, and/or the "old history" (ie an SCM that uses git could decide that the git stuff is fine for archival, and really use git as the base: and then the SCM could entirely concentrate on _only_ the "interesting" parts, ie the actual merging etc). That was really what I always personally saw "git" as, just the plumbing beneath the surface. For example, something like arch, which is based on "patches and tar-balls" (I think darcs is similar in that respect), could use git as a _hell_ of a better "history of tar-balls". The thing is, unless you take the git object database approach, using _just_ the index part doesn't really mean all that much. Sure, you could just keep the "current objects" in the object database, but quite frankly, there would probably not be a whole lot of point to that. You'd waste so much time pruning and synchronizing with your "real" database that I suspect you'd be better off not using it. (Or you could prune nightly or something, I guess). Linus - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
On Sat, Apr 16, 2005 at 03:43:02PM -0700, Linus Torvalds wrote: > On Sat, 16 Apr 2005, David Roundy wrote: > > 1) Would this actually be a good idea? It seems good to me, but there may > > be other considerations that I haven't thought of. > > I really don't know how well the git index file will work with darcs, and > the main issue is that the index file names the "stable copy" using the > sha1 hash. If darcs uses something else (and I imagine it does) you'd > need to do a fair amount of surgery, and I suspect merging changes won't > be very easy. Oh, I'm starting to see (having just browsed the git code for another half hour or so)... I had been under the (false) impression that the index file stored the contents of the files themselves, which in retrospect doesn't make any sense. So when you run update-cache --add, the file data itself immediately goes into its final hashed location, and only the sha1 info goes into the index. That's all right. Darcs would only access the cached data through a git-caching layer, and we've already got an abstraction layer over the pristine cache. As long as the git layer can quickly retrieve the contents of a given file, we should be fine. The sha1 file and tree hashing isn't direcly useful for darcs, but people will want to interoperate with git, and for that it would be nice to be able to know what the hash of a given version is. I imagine something like darcs tag --git which would tag the current version with its git hash. Of course, to implement that we only need to reproduce your algorithm for hashing trees, which probably would be easier to do ourselves without using any git code... but it would be far faster to recompute with the git backend, since git stores the hashes of all the unmodified files, and since I also imagine darcs record --git which would record a change, and then tag the resulting tree with a git hash, we might be recomputing the git hashes reasonably often, and we certainly don't want to rehash the entire kernel each time! :) > So it might well make sense to wait a bit, until the git thing has calmed > down some more. For example, I made some rather large changes > (conceptually, if not in layout of the physical file) to the index file > just yesterday, since git now uses it for merging too. > > In git, the index file isn't just a speedup, it's the "work" file _and_ > the merge entity. It's not just a floor wax, it's a dessert topping too! I think that sounds like a pretty reasonable match. In darcs, there are internally two main datatypes. One is the Patch (as you might imagine), and the other is called a "Slurpy", which is basically a tree lazily "slurped" into memory. The pristine cache is then just a way of storing the tree and so we can "slurp" it again later to retrieve the current state. So in a sense we'd be using only one side of the index file interface, the "working directory" side, where you check files out and add files in--treating it as an fast filesystem with a few extra-fancy features (like storing inodes of the files in the working directory). > I think libgit might make sense, but again, not quite yet. Maybe the new > merge model was my last smart thought even on the subject of SCM's (I kind > of hope so), but maybe it's not. > > My gut _feel_ is that the basic git low-level architecture is done, and > you can certainly start looking around and see if it matches darcs at all. Sounds good. That's sort of the feel I had gotten from other people's responses as well. We'll definitely look into how we can use (and interface with) git. -- David Roundy http://www.darcs.net - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
On Sat, 16 Apr 2005, David Roundy wrote: > > I've been thinking about the possibility of using the git "current > directory cache" code in darcs. Go wild. The license is GPLv2, with the limitation that I really do want to see v3 before I re-license anything at all, so if you take it into darcs, you'd need to add that as a per-file comment (I just doing it in the LICENSE file - I hate cluttering up individual files with tons of commentary). > So my questions are: > > 1) Would this actually be a good idea? It seems good to me, but there may > be other considerations that I haven't thought of. I really don't know how well the git index file will work with darcs, and the main issue is that the index file names the "stable copy" using the sha1 hash. If darcs uses something else (and I imagine it does) you'd need to do a fair amount of surgery, and I suspect merging changes won't be very easy. So it might well make sense to wait a bit, until the git thing has calmed down some more. For example, I made some rather large changes (conceptually, if not in layout of the physical file) to the index file just yesterday, since git now uses it for merging too. In git, the index file isn't just a speedup, it's the "work" file _and_ the merge entity. It's not just a floor wax, it's a dessert topping too! > 2) Will a license be chosen soon for git? Or has one been chosen, and I > missed it? I can't really include git code in darcs without a license. I'd > prefer GPLv2 or later (since that's how darcs is licensed), but as long as > it's at least compabible with GPLv2, I'll be all right. Yup, GPL, with the same "v2 by default" that the kernel uses). > 3) Is it likely that git will switch to not using global variables for > active_cache, active_nr and active_alloc? I wouldn't hate it, although for the intent of git, the global approach actually makes sense (dammit, I want the basic plumbing to be so small that trying to abstract it out more is a waste of time). There's simply not a lot of code that should even work at that level. But if you wait a while, and bide your time, and then spring a clean patch on me, I don't see any reason to be difficult about it either. > 4) Would there be interest in creating a libgit? I've been imagining taking > git source files and including them directly in darcs' code, but in the > long run it would be easier if there were a standard git API we could use. I think libgit might make sense, but again, not quite yet. Maybe the new merge model was my last smart thought even on the subject of SCM's (I kind of hope so), but maybe it's not. My gut _feel_ is that the basic git low-level architecture is done, and you can certainly start looking around and see if it matches darcs at all. Linus - To unsubscribe from this list: send the line "unsubscribe git" in the body of a message to [EMAIL PROTECTED] More majordomo info at http://vger.kernel.org/majordomo-info.html
Re: using git directory cache code in darcs?
* David Roundy <[EMAIL PROTECTED]> wrote: > 2) Will a license be chosen soon for git? Or has one been chosen, and > I missed it? I can't really include git code in darcs without a > license. I'd prefer GPLv2 or later (since that's how darcs is > licensed), but as long as it's at least compabible with GPLv2, I'll be > all right. there's a license in the latest code, it's GPLv2. Here's the COPYING file: --- Note that the only valid version of the GPL as far as this project is concerned is _this_ particular version of the license (ie v2, not v2.2 or v3.x or whatever), unless explicitly otherwise stated. HOWEVER, in order to allow a migration to GPLv3 if that seems like a good idea, I also ask that people involved with the project make their preferences known. In particular, if you trust me to make that decision, you might note so in your copyright message, ie something like This file is licensed under the GPL v2, or a later version at the discretion of Linus. might avoid issues. But we can also just decide to synchronize and contact all copyright holders on record if/when the occasion arises. Linus Torvalds GNU GENERAL PUBLIC LICENSE Version 2, June 1991 Copyright (C) 1989, 1991 Free Software Foundation, Inc. 59 Temple Place, Suite 330, Boston, MA 02111-1307 USA Everyone is permitted to copy and distribute verbatim copies of this license document, but changing it is not allowed. Preamble The licenses for most software are designed to take away your freedom to share and change it. By contrast, the GNU General Public License is intended to guarantee your freedom to share and change free software--to make sure the software is free for all its users. This General Public License applies to most of the Free Software Foundation's software and to any other program whose authors commit to using it. (Some other Free Software Foundation software is covered by the GNU Library General Public License instead.) You can apply it to your programs, too. When we speak of free software, we are referring to freedom, not price. Our General Public Licenses are designed to make sure that you have the freedom to distribute copies of free software (and charge for this service if you wish), that you receive source code or can get it if you want it, that you can change the software or use pieces of it in new free programs; and that you know you can do these things. To protect your rights, we need to make restrictions that forbid anyone to deny you these rights or to ask you to surrender the rights. These restrictions translate to certain responsibilities for you if you distribute copies of the software, or if you modify it. For example, if you distribute copies of such a program, whether gratis or for a fee, you must give the recipients all the rights that you have. You must make sure that they, too, receive or can get the source code. And you must show them these terms so they know their rights. We protect your rights with two steps: (1) copyright the software, and (2) offer you this license which gives you legal permission to copy, distribute and/or modify the software. Also, for each author's protection and ours, we want to make certain that everyone understands that there is no warranty for this free software. If the software is modified by someone else and passed on, we want its recipients to know that what they have is not the original, so that any problems introduced by others will not reflect on the original authors' reputations. Finally, any free program is threatened constantly by software patents. We wish to avoid the danger that redistributors of a free program will individually obtain patent licenses, in effect making the program proprietary. To prevent this, we have made it clear that any patent must be licensed for everyone's free use or not licensed at all. The precise terms and conditions for copying, distribution and modification follow. GNU GENERAL PUBLIC LICENSE TERMS AND CONDITIONS FOR COPYING, DISTRIBUTION AND MODIFICATION 0. This License applies to any program or other work which contains a notice placed by the copyright holder saying it may be distributed under the terms of this General Public License. The "Program", below, refers to any such program or work, and a "work based on the Program" means either the Program or any derivative work under copyright law: that is to say, a work containing the Program or a portion of it, either verbatim or with modifications and/or translated into another language. (Hereinafter, translation is included without limitation in the term "modification".) Each licensee is addressed as "you". Activities other than copying, distribution and modification are not covered by this License; they are outsid
using git directory cache code in darcs?
Hello Linus and various git developers (ccing darcs developers),
I've been thinking about the possibility of using the git "current
directory cache" code in darcs. Darcs already has an abstraction layer
over its pristine directory cache, so this shouldn't be too hard--provided
the git code is understandable. The default in darcs is currently to use
an actual directory ("_darcs/current/") as the cache, and we synchronize
the file modification times in the cache with those of identical files in
the working directory to speed up compares. We (the darcs developers) have
talked for some time about introducing a single-file directory cache, but
noone ever got around to it, partly because there wasn't a particularly
compelling need.
It seems that the git directory cache is precisely what we want. Also, if
we switch to (optionally) using the git directory cache, I imagine it'll
make interfacing with git a lot easier. And, of course, it would
significantly speed up a number of darcs commands, which are limited by the
slowness of the readdir-related code. We haven't tracked this down why
this is, but a recursive directory compare in which we readdir only one of
the directories (since we don't care about new files in the other one)
takes half the time of a compare in which we readdir both directories.
So my questions are:
1) Would this actually be a good idea? It seems good to me, but there may
be other considerations that I haven't thought of.
2) Will a license be chosen soon for git? Or has one been chosen, and I
missed it? I can't really include git code in darcs without a license. I'd
prefer GPLv2 or later (since that's how darcs is licensed), but as long as
it's at least compabible with GPLv2, I'll be all right.
3) Is it likely that git will switch to not using global variables for
active_cache, active_nr and active_alloc? I'd be more comfortable passing
these things around, since it would make the haskell interface easier and
safer. e.g. I'd like
struct git_cache {
struct cache_entry **cache;
unsigned int nr, alloc;
};
git_cache *read_cache(char *path_to_index);
or alternatively
int read_cache(char *path_to_index, git_cache *);
Would anyone on the git side be interested in making such changes? If not,
would they be likely to be accepted if a darcs person submitted patches?
4) Would there be interest in creating a libgit? I've been imagining taking
git source files and including them directly in darcs' code, but in the
long run it would be easier if there were a standard git API we could use.
I guess that's about all.
--
David Roundy
http://www.darcs.net
-
To unsubscribe from this list: send the line "unsubscribe git" in
the body of a message to [EMAIL PROTECTED]
More majordomo info at http://vger.kernel.org/majordomo-info.html
