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 <droundy@abridgegame.org> wrote:

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 outside its scope. The act of running the Program is not restricted, and the output from the Program is covered only if its contents constitute a work based on the Program (independent of having been made by running the Program). Whether that is true depends on what the Program does.

  1. You may copy and distribute verbatim copies of the Program's
source code as you receive it, in any medium, provided that you
conspicuously and appropriately publish on each copy an appropriate
copyright notice and disclaimer of warranty; keep intact all the
notices that refer to this License and to the absence of any warranty;
and give any other recipients of the Program a copy of this License
along with the Program.

You may charge a fee for the physical act of transferring a copy, and you may at your option offer warranty protection in exchange for a fee.

  2. You may modify your copy or copies of the Program or any portion
of it, thus forming a work based on the Program, and copy and
distribute such modifications or work under the terms of Section 1
above, provided that you also meet all of these conditions:

    a) You must cause the modified files to carry prominent notices
    stating that you changed the files and the date of any change.

    b) You must cause any work that you distribute or publish, that in
    whole or in part contains or is derived from the Program or any
    part thereof, to be licensed as a whole at no charge to all third
    parties under the terms of this License.

    c) If the modified program normally reads commands interactively
    when run, you must cause it, when started running for such
    interactive use in the most ordinary way, to print or display an
    announcement including an appropriate copyright notice and a
    notice that there is no warranty (or else, saying that you provide
    a warranty) and that users may redistribute the program under
    these conditions, and telling the user how to view a copy of this
    License.  (Exception: if the Program itself is interactive but
    does not normally print such an announcement, your work based on
    the Program is not required to print an announcement.)

These requirements apply to the modified work as a whole.  If
identifiable sections of that work are not derived from the Program,
and can be reasonably considered independent and separate works in
themselves, then this License, and its terms, do not apply to those
sections when you distribute them as separate works.  But when you
distribute the same sections as part of a whole which is a work based
on the Program, the distribution of the whole must be on the terms of
this License, whose permissions for other licensees extend to the
entire whole, and thus to each and every part regardless of who wrote it.

Thus, it is not the intent of this section to claim rights or contest your rights to work written entirely by you; rather, the intent is to exercise the right to control the distribution of derivative or collective works based on the Program.

In addition, mere aggregation of another work not based on the Program with the Program (or with a work based on the Program) on a volume of a storage or distribution medium does not bring the other work under the scope of this License.

  3. You may copy and distribute the Program (or a work based on it,
under Section 2) in object code or executable form under the terms of
Sections 1 and 2 above provided that you also do one of the following:

    a) Accompany it with the complete corresponding machine-readable
    source code, which must be distributed under the terms of Sections
    1 and 2 above on a medium customarily used for software interchange; or,

    b) Accompany it with a written offer, valid for at least three
    years, to give any third party, for a charge no more than your
    cost of physically performing source distribution, a complete
    machine-readable copy of the corresponding source code, to be
    distributed under the terms of Sections 1 and 2 above on a medium
    customarily used for software interchange; or,

    c) Accompany it with the information you received as to the offer
    to distribute corresponding source code.  (This alternative is
    allowed only for noncommercial distribution and only if you
    received the program in object code or executable form with such
    an offer, in accord with Subsection b above.)

The source code for a work means the preferred form of the work for making modifications to it. For an executable work, complete source code means all the source code for all modules it contains, plus any associated interface definition files, plus the scripts used to control compilation and installation of the executable. However, as a special exception, the source code distributed need not include anything that is normally distributed (in either source or binary form) with the major components (compiler, kernel, and so on) of the operating system on which the executable runs, unless that component itself accompanies the executable.

If distribution of executable or object code is made by offering
access to copy from a designated place, then offering equivalent
access to copy the source code from the same place counts as
distribution of the source code, even though third parties are not
compelled to copy the source along with the object code.

  4. You may not copy, modify, sublicense, or distribute the Program
except as expressly provided under this License.  Any attempt
otherwise to copy, modify, sublicense or distribute the Program is
void, and will automatically terminate your rights under this License.
However, parties who have received copies, or rights, from you under
this License will not have their licenses terminated so long as such
parties remain in full compliance.

  5. You are not required to accept this License, since you have not
signed it.  However, nothing else grants you permission to modify or
distribute the Program or its derivative works.  These actions are
prohibited by law if you do not accept this License.  Therefore, by
modifying or distributing the Program (or any work based on the
Program), you indicate your acceptance of this License to do so, and
all its terms and conditions for copying, distributing or modifying
the Program or works based on it.

  6. Each time you redistribute the Program (or any work based on the
Program), the recipient automatically receives a license from the
original licensor to copy, distribute or modify the Program subject to
these terms and conditions.  You may not impose any further
restrictions on the recipients' exercise of the rights granted herein.
You are not responsible for enforcing compliance by third parties to
this License.

  7. If, as a consequence of a court judgment or allegation of patent
infringement or for any other reason (not limited to patent issues),
conditions are imposed on you (whether by court order, agreement or
otherwise) that contradict the conditions of this License, they do not
excuse you from the conditions of this License.  If you cannot
distribute so as to satisfy simultaneously your obligations under this
License and any other pertinent obligations, then as a consequence you
may not distribute the Program at all.  For example, if a patent
license would not permit royalty-free redistribution of the Program by
all those who receive copies directly or indirectly through you, then
the only way you could satisfy both it and this License would be to
refrain entirely from distribution of the Program.

If any portion of this section is held invalid or unenforceable under any particular circumstance, the balance of the section is intended to apply and the section as a whole is intended to apply in other circumstances.

It is not the purpose of this section to induce you to infringe any patents or other property right claims or to contest validity of any such claims; this section has the sole purpose of protecting the integrity of the free software distribution system, which is implemented by public license practices. Many people have made generous contributions to the wide range of software distributed through that system in reliance on consistent application of that system; it is up to the author/donor to decide if he or she is willing to distribute software through any other system and a licensee cannot impose that choice.

This section is intended to make thoroughly clear what is believed to
be a consequence of the rest of this License.

  8. If the distribution and/or use of the Program is restricted in
certain countries either by patents or by copyrighted interfaces, the
original copyright holder who places the Program under this License
may add an explicit geographical distribution limitation excluding
those countries, so that distribution is permitted only in or among
countries not thus excluded.  In such case, this License incorporates
the limitation as if written in the body of this License.

  9. The Free Software Foundation may publish revised and/or new versions
of the General Public License from time to time.  Such new versions will
be similar in spirit to the present version, but may differ in detail to
address new problems or concerns.

Each version is given a distinguishing version number. If the Program specifies a version number of this License which applies to it and "any later version", you have the option of following the terms and conditions either of that version or of any later version published by the Free Software Foundation. If the Program does not specify a version number of this License, you may choose any version ever published by the Free Software Foundation.

  10. If you wish to incorporate parts of the Program into other free
programs whose distribution conditions are different, write to the author
to ask for permission.  For software which is copyrighted by the Free
Software Foundation, write to the Free Software Foundation; we sometimes
make exceptions for this.  Our decision will be guided by the two goals
of preserving the free status of all derivatives of our free software and
of promoting the sharing and reuse of software generally.

			    NO WARRANTY

  11. BECAUSE THE PROGRAM IS LICENSED FREE OF CHARGE, THERE IS NO WARRANTY
FOR THE PROGRAM, TO THE EXTENT PERMITTED BY APPLICABLE LAW.  EXCEPT WHEN
OTHERWISE STATED IN WRITING THE COPYRIGHT HOLDERS AND/OR OTHER PARTIES
PROVIDE THE PROGRAM "AS IS" WITHOUT WARRANTY OF ANY KIND, EITHER EXPRESSED
OR IMPLIED, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF
MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE.  THE ENTIRE RISK AS
TO THE QUALITY AND PERFORMANCE OF THE PROGRAM IS WITH YOU.  SHOULD THE
PROGRAM PROVE DEFECTIVE, YOU ASSUME THE COST OF ALL NECESSARY SERVICING,
REPAIR OR CORRECTION.

  12. IN NO EVENT UNLESS REQUIRED BY APPLICABLE LAW OR AGREED TO IN WRITING
WILL ANY COPYRIGHT HOLDER, OR ANY OTHER PARTY WHO MAY MODIFY AND/OR
REDISTRIBUTE THE PROGRAM AS PERMITTED ABOVE, BE LIABLE TO YOU FOR DAMAGES,
INCLUDING ANY GENERAL, SPECIAL, INCIDENTAL OR CONSEQUENTIAL DAMAGES ARISING
OUT OF THE USE OR INABILITY TO USE THE PROGRAM (INCLUDING BUT NOT LIMITED
TO LOSS OF DATA OR DATA BEING RENDERED INACCURATE OR LOSSES SUSTAINED BY
YOU OR THIRD PARTIES OR A FAILURE OF THE PROGRAM TO OPERATE WITH ANY OTHER
PROGRAMS), EVEN IF SUCH HOLDER OR OTHER PARTY HAS BEEN ADVISED OF THE
POSSIBILITY OF SUCH DAMAGES.

		     END OF TERMS AND CONDITIONS

	    How to Apply These Terms to Your New Programs

  If you develop a new program, and you want it to be of the greatest
possible use to the public, the best way to achieve this is to make it
free software which everyone can redistribute and change under these terms.

  To do so, attach the following notices to the program.  It is safest
to attach them to the start of each source file to most effectively
convey the exclusion of warranty; and each file should have at least
the "copyright" line and a pointer to where the full notice is found.

    <one line to give the program's name and a brief idea of what it does.>
    Copyright (C) <year>  <name of author>

    This program is free software; you can redistribute it and/or modify
    it under the terms of the GNU General Public License as published by
    the Free Software Foundation; either version 2 of the License, or
    (at your option) any later version.

    This program is distributed in the hope that it will be useful,
    but WITHOUT ANY WARRANTY; without even the implied warranty of
    MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE.  See the
    GNU General Public License for more details.

    You should have received a copy of the GNU General Public License
    along with this program; if not, write to the Free Software
    Foundation, Inc., 59 Temple Place, Suite 330, Boston, MA  02111-1307  USA

Also add information on how to contact you by electronic and paper mail.

If the program is interactive, make it output a short notice like this when it starts in an interactive mode:

    Gnomovision version 69, Copyright (C) year name of author
    Gnomovision comes with ABSOLUTELY NO WARRANTY; for details type `show w'.
    This is free software, and you are welcome to redistribute it
    under certain conditions; type `show c' for details.

The hypothetical commands `show w' and `show c' should show the appropriate parts of the General Public License. Of course, the commands you use may be called something other than `show w' and `show c'; they could even be mouse-clicks or menu items--whatever suits your program.

You should also get your employer (if you work as a programmer) or your school, if any, to sign a "copyright disclaimer" for the program, if necessary. Here is a sample; alter the names:

  Yoyodyne, Inc., hereby disclaims all copyright interest in the program
  `Gnomovision' (which makes passes at compilers) written by James Hacker.

  <signature of Ty Coon>, 1 April 1989
  Ty Coon, President of Vice

This General Public License does not permit incorporating your program into proprietary programs. If your program is a subroutine library, you may consider it more useful to permit linking proprietary applications with the library. If this is what you want to do, use the GNU Library General Public License instead of this License.

DR> 1) Would this actually be a good idea?

I think it is sensible, especially if you are doing a lot of comparison between the working area and the pristine.

DR> 3) Is it likely that git will switch to not using global DR> variables for active_cache, active_nr and active_alloc?

DR> 4) Would there be interest in creating a libgit?

These are related. I have seen some people interested in libifying it, and encapsulating those globals would naturally fall out of it. My impression from the list however is that a lot more people are interested in the upper SCM layer than the git layer right now. And git layer, although solid enough to host itself, is still slushy. A couple of days ago dircache format was changed from host to network endian. Last night Linus made another change to dircache format, which fortunately is upward compatible if you stay within pathnames shorter than 2^12 bytes ;-). Another problem I see for somebody to pick up and start libifying things right now is that, although there is one central person on the SCM side (Petr Baudis), git layer is still fractured between Linus and Petr.

Petr syncs with Linus often and he seems to be doing a good job at keeping track of public patches, but the git layer Linus works on does not have some patches Petr collected or wrote himself. In time, a better coordination would emerge, of course, but the project is still young at this moment. Stay tuned ;-).

On Sat, 16 Apr 2005, David Roundy wrote:

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).

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!

Yup, GPL, with the same "v2 by default" that the kernel uses).

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.

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

On Sat, Apr 16, 2005 at 03:43:02PM -0700, Linus Torvalds wrote:

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! :)

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).

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.

On Sun, 17 Apr 2005, David Roundy wrote:

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

Linus Torvalds wrote:

If you don't want it, I won't do it. Still makes sense to separate the plumbing from the porcelain, though.

Linus Torvalds schrieb:

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

NA> Linus Torvalds schrieb:

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.

Junio wrote:

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.

Linus wrote:

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.

On Sun, 17 Apr 2005, Paul Jackson wrote:

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

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.

Would it be useful at this point to make common and centralize some/most of the various options that control git? (as well as add some useful ones). Something like:

struct _git_opt {
         int verbose:1;
         int debug:1;
	int dry-run:1;
         int should_block:1;
         int remove_lock:1;
         int allow_add:1;
         int allow_remove:1;
         int null_termination:1;
         int show_cached:1;
         int show_deleted:1;
         int show_others:1;
         int show_ignored:1;
         int show_stage:1;
         int show_unmerged:1;
         int show_edges:1;
         int show_unreachable:1;
         int basemask:1;
         int recursive:1;
         int force_filename:1;
         int force:1;
         int quiet:1;
         int stage:1;
};