<!DOCTYPE html PUBLIC "-//W3C//DTD XHTML 1.0 Strict//EN" "http://www.w3.org/TR/xhtml1/DTD/xhtml1-strict.dtd"> <html xmlns="http://www.w3.org/1999/xhtml"> <head> <style type="text/css"> /* <![CDATA[ */ @import "branding/css/tigris.css"; @import "branding/css/inst.css"; /* ]]> */</style> <link rel="stylesheet" type="text/css" media="print" href="branding/css/print.css" /> <script type="text/javascript" src="branding/scripts/tigris.js"></script> <title>Subversion Design</title> </head> <body> <div class="app"> <div class="h1"> <h1 style="text-align: center">Subversion Design</h1> </div> <p class="warningmark"><em>NOTE: This document is out of date. The last substantial update was in October 2002 (r3377). However, people often come here for the section on the <a href="#server.fs.struct.bubble-up">directory bubble-up method</a>, which is still accurate.</em></p> <div class="h1"> <h2>Table of Contents</h2> <ol id="toc"> <li><a href="#goals">Goals — The goals of the Subversion project</a> <ol> <li><a href="#goals.rename-remove-resurrect">Rename/removal/resurrection support</a></li> <li><a href="#goals.textbinary">Text vs binary issues</a></li> <li><a href="#goals.i18n">I18N/Multilingual support</a></li> <li><a href="#goals.branching-and-tagging">Branching and tagging</a></li> <li><a href="#goals.misc">Miscellaneous new behaviors</a> <ol> <li><a href="#goals.misc.logmsgs">Log messages</a></li> <li><a href="#goals.misc.diffplugins">Client side diff plug-ins</a></li> <li><a href="#goals.misc.merging">Better merging</a></li> <li><a href="#goals.misc.conflicts">Conflicts resolution</a></li> </ol> </li> <!-- goals.misc --> </ol> </li> <!-- goals --> <li><a href="#model">Model — The versioning model used by Subversion</a> <ol> <li><a href="#model.wc-and-repos">Working Directories and Repositories</a></li> <li><a href="#model.txns-and-revnums">Transactions and Revision Numbers</a></li> <li><a href="#model.how-wc">How Working Directories Track the Repository</a></li> <li><a href="#model.lock-merge">Locking vs. Merging - Two Paradigms of Co-operative Developments</a></li> <li><a href="#model.props">Properties</a></li> <li><a href="#model.merging-and-ancestry">Merging and Ancestry</a></li> </ol> </li> <!-- model --> <li><a href="#archi">Architecture — How Subversion's components work together</a> <ol> <li><a href="#archi.client">Client Layer</a></li> <li><a href="#archi.network">Network Layer</a></li> <li><a href="#archi.fs">Filesystem Layer</a></li> </ol> </li> <!-- archi --> <li><a href="#deltas">Deltas — How to describe changes</a> <ol> <li><a href="#deltas.text">Text Deltas</a></li> <li><a href="#deltas.prop">Property Deltas</a></li> <li><a href="#deltas.tree">Tree Deltas</a></li> <li><a href="#deltas.postfix-text">Postfix Text Deltas</a></li> <li><a href="#deltas.serializing-via-editor">Serializing Deltas via the "Editor" Interface</a></li> </ol> </li> <!-- deltas --> <li><a href="#client">Client — How the client works</a> <ol> <li><a href="#client.wc">Working copies and the working copy library</a> <ol> <li><a href="#client.wc.layout">The layout of working copies</a></li> <li><a href="#client.wc.library">The working copy management library</a></li> </ol> </li> <!-- client.wc --> <li><a href="#client.libsvn_ra">The repository access library</a></li> <li><a href="#client.libsvn_client">The client operation library</a></li> </ol> </li> <!-- client --> <li><a href="#protocol">Protocol — How the client and server communicate</a> <ol> <li><a href="#protocol.webdav">The HTTP/WebDAV/DeltaV based protocol</a></li> <li><a href="#protocol.svn">The custom protocol</a></li> </ol> </li> <!-- protocol --> <li><a href="#server">Server — How the server works</a> <ol> <li><a href="#server.fs">Filesystem</a> <ol> <li><a href="#server.fs.overview">Filesystem Overview</a></li> <li><a href="#server.fs.api">API</a></li> <li><a href="#server.fs.struct">Repository Structure</a> <ol> <li><a href="#server.fs.struct.schema">Schema</a></li> <li><a href="#server.fs.struct.bubble-up">Bubble-Up Method</a></li> <li><a href="#server.fs.struct.diffy-storage">Diffy Storage</a></li> </ol> </li> <!-- server.fs.struct --> <li><a href="#server.fs.implementation">Implementation</a></li> </ol> </li> <!-- server.fs --> <li><a href="#server.libsvn_repos">Repository Library</a></li> </ol> </li> <!-- server --> <li><a href="#license">License — Copyright</a></li> </ol> </div> <!-- ================================================================ Copyright (c) 1999-2004 CollabNet. All rights reserved. Redistribution and use in source and binary forms, with or without modification, are permitted provided that the following conditions are met: 1. Redistributions of source code must retain the above copyright notice, this list of conditions and the following disclaimer. 2. Redistributions in binary form must reproduce the above copyright notice, this list of conditions and the following disclaimer in the documentation and/or other materials provided with the distribution. 3. The end-user documentation included with the redistribution, if any, must include the following acknowledgment: "This product includes software developed by CollabNet (http://www.Collab.Net/)." Alternately, this acknowledgment may appear in the software itself, if and wherever such third-party acknowledgments normally appear. 4. The hosted project names must not be used to endorse or promote products derived from this software without prior written permission. For written permission, please contact info@collab.net. 5. Products derived from this software may not use the "Tigris" name nor may "Tigris" appear in their names without prior written permission of CollabNet. THIS SOFTWARE IS PROVIDED ``AS IS'' AND ANY EXPRESSED OR IMPLIED WARRANTIES, INCLUDING, BUT NOT LIMITED TO, THE IMPLIED WARRANTIES OF MERCHANTABILITY AND FITNESS FOR A PARTICULAR PURPOSE ARE DISCLAIMED. IN NO EVENT SHALL COLLABNET OR ITS CONTRIBUTORS BE LIABLE FOR ANY DIRECT, INDIRECT, INCIDENTAL, SPECIAL, EXEMPLARY, OR CONSEQUENTIAL DAMAGES (INCLUDING, BUT NOT LIMITED TO, PROCUREMENT OF SUBSTITUTE GOODS OR SERVICES; LOSS OF USE, DATA, OR PROFITS; OR BUSINESS INTERRUPTION) HOWEVER CAUSED AND ON ANY THEORY OF LIABILITY, WHETHER IN CONTRACT, STRICT LIABILITY, OR TORT (INCLUDING NEGLIGENCE OR OTHERWISE) ARISING IN ANY WAY OUT OF THE USE OF THIS SOFTWARE, EVEN IF ADVISED OF THE POSSIBILITY OF SUCH DAMAGE. ==================================================================== This software consists of voluntary contributions made by many individuals on behalf of CollabNet. --> <div class="h2" id="goals" title="#goals"> <h2>Goals — The goals of the Subversion project</h2> <p>The goal of the Subversion project is to write a version control system that takes over CVS's current and future user base (If you're not familiar with CVS or its shortcomings, then skip to <a href="#model">Model — The versioning model used by Subversion</a>) . The first release has all the major features of CVS, plus certain new features that CVS users often wish they had. In general, Subversion works like CVS, except where there's a compelling reason to be different.</p> <p>So what does Subversion have that CVS doesn't?</p> <ul> <li><p>It versions directories, file-metadata, renames, copies and removals/resurrections. In other words, Subversion records the changes users make to directory trees, not just changes to file contents.</p></li> <li><p>Tagging and branching are constant-time and constant-space.</p></li> <li><p>It is natively client-server, hence much more maintainable than CVS. (In CVS, the client-server protocol was added as an afterthought. This means that most new features have to be implemented twice, or at least more than once: code for the local case, and code for the client-server case.)</p></li> <li><p>The repository is organized efficiently and comprehensibly. (Without going into too much detail, let's just say that CVS's repository structure is showing its age.)</p></li> <li><p>Commits are atomic. Each commit results in a single revision number, which refers to the state of the entire tree. Files no longer have their own revision numbers.</p></li> <li><p>The locking scheme is only as strict as absolutely necessary. Reads are never locked, and writes lock only the files being written, for only as long as needed.</p></li> <li><p>It has internationalization support.</p></li> <li><p>It handles binary files gracefully (experience has shown that CVS's binary file handling is prone to user error).</p></li> <li><p>It takes advantage of the Net's experience with CVS by choosing better default behaviors for certain situations.</p></li> </ul> <p>Some of these advantages are clear and require no further discussion. Others are not so obvious, and are explained in greater detail below.</p> <div class="h3" id="goals.rename-remove-resurrect" title="#goals.rename-remove-resurrect"> <h3>Rename/removal/resurrection support</h3> <p>Full rename support means you can trace through ancestry by name <em>or</em> by entity. For example, if you say "Give me revision 12 of foo.c", do you mean revision 12 of the file whose name is <em>now</em> foo.c (but perhaps it was named bar.c back at revision 12), or the file whose name was foo.c in revision 12 (perhaps that file no longer exists, or has a different name now)? In Subversion, both interpretations are available to the user.</p> <p>(Note: we've not yet implemented this, but it wouldn't be too hard. People are advocating switches to 'svn log' that cause history to be traced backwards either by entity or by path.)</p> </div> <!-- goals.rename-remove-resurrect (h3) --> <div class="h3" id="goals.textbinary" title="#goals.textbinary"> <h3>Text vs binary issues</h3> <p>Historically, binary files have been problematic in CVS for two unrelated reasons: keyword expansion, and line-end conversion.</p> <ul> <li><p><strong class="firstterm">Keyword expansion</strong> is when CVS expands "$Revision$" into "$Revision: 1.1 $", for example. There are a number of keywords in CVS: "$Author: sussman $", "$Date: 2001/06/04 22:00:52 $", and so on.</p></li> <li><p><strong class="firstterm">Line-end conversion</strong> is when CVS gives plaintext files the appropriate line-ending conventions for the working copy's platform. For example, Unix working copies use LF, but Windows working copies use CRLF. (Like CVS, the Subversion repository stores text files in Unix LF format).</p></li> </ul> <p>Both keyword substitution and line-end conversion are sensible only for plain text files. CVS only recognizes two file types anyway: plaintext and binary. And CVS assumes files are plain text unless you tell it otherwise.</p> <p>Subversion recognizes the same two types. The question is, how does it determine a file's type? Experience with CVS suggests that assuming text unless told otherwise is a losing strategy – people frequently forget to mark images and other opaque formats as binary, then later they wonder why CVS mangled their data. So Subversion will not mangle data: when moving over the network, or when being stored in the repository, it treats all files as binary. In the working copy, a tweakable meta-data property indicates whether to treat the file as text or binary for purposes of whether or not to allow contextual merging during updates.</p> <p>Users can turn line-end conversion on or off per file by tweaking meta-data. Files do <em>not</em> undergo keyword substitution by default, on the theory that if someone wants substitution and isn't getting it, they'll look in the manual; but if they are getting it and didn't want it, they might just be confused and not know what to do. Users can turn substitution on or off per file.</p> <p>Both of these changes are done on the client side; the repository does not even know about them.</p> </div> <!-- goals.textbinary (h3) --> <div class="h3" id="goals.i18n" title="#goals.i18n"> <h3>I18N/Multilingual support</h3> <p>Subversion is internationalized – commands, user messages, and errors can be customized to the appropriate human language at build-time (or run time, if that's not much harder).</p> <p>File names and contents may be multilingual; Subversion does not assume an ASCII-only universe. For purposes of keyword expansion and line-end conversion, Subversion also understands the UTF-* encodings (but not necessarily all of them by the first release).</p> </div> <!-- goals.i18n (h3) --> <div class="h3" id="goals.branching-and-tagging" title="#goals.branching-and-tagging"> <h3>Branching and tagging</h3> <p>Subversion supports branching and tagging with one efficient operation: `clone'. To clone a tree is to copy it, to create another tree exactly like it (except that the new tree knows its ancestry relationship to the old one).</p> <p>At the moment of creation, a clone requires only a small, constant amount of space in the repository – most of its storage is shared with the original tree. If you never commit anything on the clone, then it's just like a CVS tag. If you start committing on it, then it's a branch. Voila! This also implies CVS's "vendor branching" feature, since Subversion has real rename and directory support.</p> </div> <!-- goals.branching-and-tagging (h3) --> <div class="h3" id="goals.misc" title="#goals.misc"> <h3>Miscellaneous new behaviors</h3> <div class="h4" id="goals.misc.logmsgs" title="#goals.misc.logmsgs"> <h4>Log messages</h4> <p>Subversion has a flexible log message policy (a small matter, but one dear to our hearts).</p> <p>Log messages should be a matter of project policy, not version control software policy. If a user commits with no log message, then Subversion defaults to an empty message. (CVS tries to require log messages, but fails: we've all seen empty log messages in CVS, where the user committed with deliberately empty quotes. Let's stop the madness now.)</p> </div> <!-- goals.misc.logmsgs (h4) --> <div class="h4" id="goals.misc.diffplugins" title="#goals.misc.diffplugins"> <h4>Client side diff plug-ins</h4> <p>Subversion supports client-side plug-in diff programs.</p> <p>There is no need for Subversion to have every possible diff mechanism built in. It can invoke a user-specified client-side diff program on the two revisions of the file(s) locally.</p> <p>(Note: This feature does not exist yet, but is planned for post-1.0.)</p> </div> <!-- goals.misc.diffplugins (h4) --> <div class="h4" id="goals.misc.merging" title="#goals.misc.merging"> <h4>Better merging</h4> <p>Subversion remembers what has already been merged in and what hasn't, thereby avoiding the problem, familiar to CVS users, of spurious conflicts on repeated merges.</p> <p>(Note: Parts of his feature (<a href="/merge-tracking/">Merge Tracking</a>) are implemented in Subversion 1.5; see the <a href="svn_1.5_releasenotes.html#merge-tracking" >release notes</a>.)</p> <p>For details, see <a href="#model.merging-and-ancestry">Merging and Ancestry</a>.</p> </div> <!-- goals.misc.merging (h4) --> <div class="h4" id="goals.misc.conflicts" title="#goals.misc.conflicts"> <h4>Conflicts resolution</h4> <p>For text files, Subversion resolves conflicts similarly to CVS, by folding repository changes into the working files with conflict markers. But, for <em>both</em> text and binary files, Subversion also always puts the old and new pristine repository revisions into temporary files, and the pristine working copy revision in another temporary file.</p> <p>Thus, for any conflict, the user has four files readily at hand:</p> <ol> <li><p>the original working copy file with local mods</p></li> <li><p>the older repository file</p></li> <li><p>the newest repository file</p></li> <li><p>the merged file, with conflict markers</p></li> </ol> <p>and in a binary file conflict, the user has all but the last.</p> <p>When the conflict has been resolved and the working copy is committed, Subversion automatically removes the temporary pristine files.</p> <p>A more general solution would allow plug-in merge resolution tools on the client side; but this is not scheduled for the first release). Note that users can use their own merge tools anyway, since all the original files are available.</p> </div> <!-- goals.misc.conflicts (h4) --> </div> <!-- goals.misc (h3) --> </div> <!-- goals (h2) --> <div class="h2" id="model" title="#model"> <h2>Model — The versioning model used by Subversion</h2> <p>This chapter explains the user's view of Subversion — what “objects” you interact with, how they behave, and how they relate to each other.</p> <div class="h3" id="model.wc-and-repos" title="#model.wc-and-repos"> <h3>Working Directories and Repositories</h3> <p>Suppose you are using Subversion to manage a software project. There are two things you will interact with: your working directory, and the repository.</p> <p>Your <strong class="firstterm">working directory</strong> is an ordinary directory tree, on your local system, containing your project's sources. You can edit these files and compile your program from them in the usual way. Your working directory is your own private work area: Subversion never changes the files in your working directory, or publishes the changes you make there, until you explicitly tell it to do so.</p> <p>After you've made some changes to the files in your working directory, and verified that they work properly, Subversion provides commands to publish your changes to the other people working with you on your project. If they publish their own changes, Subversion provides commands to incorporate those changes into your working directory.</p> <p>A working directory contains some extra files, created and maintained by Subversion, to help it carry out these commands. In particular, these files help Subversion recognize which files contain unpublished changes, and which files are out-of-date with respect to others' work.</p> <p>While your working directory is for your use alone, the <strong class="firstterm">repository</strong> is the common public record you share with everyone else working on the project. To publish your changes, you use Subversion to put them in the repository. (What this means, exactly, we explain below.) Once your changes are in the repository, others can tell Subversion to incorporate your changes into their working directories. In a collaborative environment like this, each user will typically have their own working directory (or perhaps more than one), and all the working directories will be backed by a single repository, shared amongst all the users.</p> <p>A Subversion repository holds a single directory tree, and records the history of changes to that tree. The repository retains enough information to recreate any prior state of the tree, compute the differences between any two prior trees, and report the relations between files in the tree — which files are derived from which other files.</p> <p>A Subversion repository can hold the source code for several projects; usually, each project is a subdirectory in the tree. In this arrangement, a working directory will usually correspond to a particular subtree of the repository.</p> <p>For example, suppose you have a repository laid out like this:</p> <pre> /trunk/paint/Makefile canvas.c brush.c write/Makefile document.c search.c </pre> <p>In other words, the repository's root directory has a single subdirectory named <tt class="filename">trunk</tt>, which itself contains two subdirectories: <tt class="filename">paint</tt> and <tt class="filename">write</tt>.</p> <p>To get a working directory, you must <strong class="firstterm">check out</strong> some subtree of the repository. If you check out <tt class="filename">/trunk/write</tt>, you will get a working directory like this:</p> <pre> write/Makefile document.c search.c .svn/ </pre> <p>This working directory is a copy of the repository's <tt class="filename">/trunk/write</tt> directory, with one additional entry — <tt class="filename">.svn</tt> — which holds the extra information needed by Subversion, as mentioned above.</p> <p>Suppose you make changes to <tt class="filename">search.c</tt>. Since the <tt class="filename">.svn</tt> directory remembers the file's modification date and original contents, Subversion can tell that you've changed the file. However, Subversion does not make your changes public until you explicitly tell it to.</p> <p>To publish your changes, you can use Subversion's ‘<tt class="literal">commit</tt>’ command:</p> <pre> $ pwd /home/jimb/write $ ls -a .svn/ Makefile document.c search.c $ svn commit search.c $ </pre> <p>Now your changes to <tt class="filename">search.c</tt> have been committed to the repository; if another user checks out a working copy of <tt class="filename">/trunk/write</tt>, they will see your text.</p> <p>Suppose you have a collaborator, Felix, who checked out a working directory of <tt class="filename">/trunk/write</tt> at the same time you did. When you commit your change to <tt class="filename">search.c</tt>, Felix's working copy is left unchanged; Subversion only modifies working directories at the user's request.</p> <p>To bring his working directory up to date, Felix can use the Subversion ‘<tt class="literal">update</tt>’ command. This will incorporate your changes into his working directory, as well as any others that have been committed since he checked it out.</p> <pre> $ pwd /home/felix/write $ ls -a .svn/ Makefile document.c search.c $ svn update U search.c $ </pre> <p>The output from the ‘<tt class="literal">svn update</tt>’ command indicates that Subversion updated the contents of <tt class="filename">search.c</tt>. Note that Felix didn't need to specify which files to update; Subversion uses the information in the <tt class="filename">.svn</tt> directory, and further information in the repository, to decide which files need to be brought up to date.</p> <p>We explain below what happens when both you and Felix make changes to the same file.</p> </div> <!-- model.wc-and-repos (h3) --> <div class="h3" id="model.txns-and-revnums" title="#model.txns-and-revnums"> <h3>Transactions and Revision Numbers</h3> <p>A Subversion ‘<tt class="literal">commit</tt>’ operation can publish changes to any number of files and directories as a single atomic transaction. In your working directory, you can change files' contents, create, delete, rename and copy files and directories, and then commit the completed set of changes as a unit.</p> <p>In the repository, each commit is treated as an atomic transaction: either all the commit's changes take place, or none of them take place. Subversion tries to retain this atomicity in the face of program crashes, system crashes, network problems, and other users' actions. We may call a commit a <strong class="firstterm">transaction</strong> when we want to emphasize its indivisible nature.</p> <p>Each time the repository accepts a transaction, this creates a new state of the tree, called a <strong class="firstterm">revision</strong>. Each revision is assigned a unique natural number, one greater than the number of the previous revision. The initial revision of a freshly created repository is numbered zero, and consists of an empty root directory.</p> <p>Since each transaction creates a new revision, with its own number, we can also use these numbers to refer to transactions; transaction <em class="replaceable">n</em> is the transaction which created revision <em class="replaceable">n</em>. There is no transaction numbered zero.</p> <p>Unlike those of many other systems, Subversion's revision numbers apply to an entire tree, not individual files. Each revision number selects an entire tree.</p> <p>It's important to note that working directories do not always correspond to any single revision in the repository; they may contain files from several different revisions. For example, suppose you check out a working directory from a repository whose most recent revision is 4:</p> <pre> write/Makefile:4 document.c:4 search.c:4 </pre> <p>At the moment, this working directory corresponds exactly to revision 4 in the repository. However, suppose you make a change to <tt class="filename">search.c</tt>, and commit that change. Assuming no other commits have taken place, your commit will create revision 5 of the repository, and your working directory will look like this:</p> <pre> write/Makefile:4 document.c:4 search.c:5 </pre> <p>Suppose that, at this point, Felix commits a change to <tt class="filename">document.c</tt>, creating revision 6. If you use ‘<tt class="literal">svn update</tt>’ to bring your working directory up to date, then it will look like this:</p> <pre> write/Makefile:6 document.c:6 search.c:6 </pre> <p>Felix's changes to <tt class="filename">document.c</tt> will appear in your working copy of that file, and your change will still be present in <tt class="filename">search.c</tt>. In this example, the text of <tt class="filename">Makefile</tt> is identical in revisions 4, 5, and 6, but Subversion will mark your working copy with revision 6 to indicate that it is still current. So, after you do a clean update at the root of your working directory, your working directory will generally correspond exactly to some revision in the repository.</p> </div> <!-- model.txns-and-revnums (h3) --> <div class="h3" id="model.how-wc" title="#model.how-wc"> <h3>How Working Directories Track the Repository</h3> <p>For each file in a working directory, Subversion records two essential pieces of information:</p> <ul> <li><p>what revision of what repository file your working copy is based on (this is called the file's <strong class="firstterm">base revision</strong>), and</p></li> <li><p>a timestamp recording when the local copy was last updated.</p></li> </ul> <p>Given this information, by talking to the repository, Subversion can tell which of the following four states a file is in:</p> <ul> <li><p><strong>Unchanged, and current.</strong> The file is unchanged in the working directory, and no changes to that file have been committed to the repository since its base revision.</p></li> <li><p><strong>Locally changed, and current</strong>. The file has been changed in the working directory, and no changes to that file have been committed to the repository since its base revision. There are local changes that have not been committed to the repository.</p></li> <li><p><strong>Unchanged, and out-of-date</strong>. The file has not been changed in the working directory, but it has been changed in the repository. The file should eventually be updated, to make it current with the public revision.</p></li> <li><p><strong>Locally changed, and out-of-date</strong>. The file has been changed both in the working directory, and in the repository. The file should be updated; Subversion will attempt to merge the public changes with the local changes. If it can't complete the merge in a plausible way automatically, Subversion leaves it to the user to resolve the conflict.</p></li> </ul> </div> <!-- model.how-wc (h3) --> <div class="h3" id="model.lock-merge" title="#model.lock-merge"> <h3>Locking vs. Merging - Two Paradigms of Co-operative Developments</h3> <p>By default, Subversion prefers the “merging” method of handling simultaneous editing by multiple users. This means that Subversion does not prevent two users from making changes to the same file at the same time. For example, if both you and Felix have checked out working directories of <tt class="filename">/trunk/write</tt>, Subversion will allow both of you to change <tt class="filename">write/search.c</tt> in your working directories. Then, the following sequence of events will occur:</p> <ul> <li><p>Suppose Felix tries to commit his changes to <tt class="filename">search.c</tt> first. His commit will succeed, and his text will appear in the latest revision in the repository.</p></li> <li><p>When you attempt to commit your changes to <tt class="filename">search.c</tt>, Subversion will reject your commit, and tell you that you must update <tt class="filename">search.c</tt> before you can commit it.</p></li> <li><p>When you update <tt class="filename">search.c</tt>, Subversion will try to merge Felix's changes from the repository with your local changes. By default, Subversion merges as if it were applying a patch: if your local changes do not overlap textually with Felix's, then all is well; otherwise, Subversion leaves it to you to resolve the overlapping changes. In either case, Subversion carefully preserves a copy of the original pre-merge text.</p></li> <li><p>Once you have verified that Felix's changes and your changes have been merged correctly, you can commit the new revision of <tt class="filename">search.c</tt>, which now contains everyone's changes.</p></li> </ul> <p>Some version control systems provide “locks”, which prevent others from changing a file once one person has begun working on it. In our experience, merging is preferable to locks, because:</p> <ul> <li><p>changes usually do not conflict, so Subversion's behavior does the right thing by default, while locking can interfere with legitimate work;</p></li> <li><p>locking can prevent conflicts within a file, but not conflicts between files (say, between a C header file and another file that includes it), so it doesn't really solve the problem; and finally,</p></li> <li><p>people often forget that they are holding locks, resulting in unnecessary delays and friction.</p></li> </ul> <p>Of course, some kinds of files with rigid formats, like images or executables, are simply not mergeable. To support this, Subversion allows users to customize its merging behavior on a per-file basis. Firstly, you can direct Subversion to refuse to merge changes to certain files, and simply present you with the two original texts to choose from. Secondly, in Subversion 1.2 and later, support for the “locking” method of working is also available, and individual files can be designated as requiring locking.</p> <p>(In the future, you may be able to direct Subversion to merge using a tool which respects the semantics of specific complex file formats.)</p> </div> <!-- model.lock-merge (h3) --> <div class="h3" id="model.props" title="#model.props"> <h3>Properties</h3> <p>Files generally have interesting attributes beyond their contents: mime-types, executable permissions, EOL styles, and so on. Subversion attempts to preserve these attributes, or at least record them, when doing so would be meaningful. However, different operating systems support very different sets of file attributes: Windows NT supports access control lists, while Linux provides only the simpler traditional Unix permission bits.</p> <p>In order to interoperate well with clients on many different operating systems, Subversion supports <strong class="firstterm">property lists</strong>, a simple, general-purpose mechanism which clients can use to store arbitrary out-of-band information about files.</p> <p>A property list is a set of name / value pairs. A property name is an arbitrary text string, expressed as a Unicode UTF-8 string, canonically decomposed and ordered. A property value is an arbitrary string of bytes. Property values may be of any size, but Subversion may not handle very large property values efficiently. No two properties in a given a property list may have the same name. Although the word `list' usually denotes an ordered sequence, there is no fixed order to the properties in a property list; the term `property list' is historical.</p> <p>Each revision number, file, directory, and directory entry in the Subversion repository, has its own property list. Subversion puts these property lists to several uses:</p> <ul> <li><p>Clients can use properties to store file attributes, as described above.</p></li> <li><p>The Subversion server uses properties to hold attributes of its own, and allow clients to read and modify them. For example, someday a hypothetical ‘<tt class="literal">svn-acl</tt>’ property might hold an access control list which the Subversion server uses to regulate access to repository files.</p></li> <li><p>Users can invent properties of their own, to store arbitrary information for use by scripts, build environments, and so on. Names of user properties should be URI's, to avoid conflicts between organizations.</p></li> </ul> <p>Property lists are versioned, just like file contents. You can change properties in your working directory, but those changes are not visible in the repository until you commit your local changes. If you do commit a change to a property value, other users will see your change when they update their working directories.</p> </div> <!-- model.props (h3) --> <div class="h3" id="model.merging-and-ancestry" title="#model.merging-and-ancestry"> <h3>Merging and Ancestry</h3> <p>[WARNING: this section was written in May 2000, at the very beginning of the Subversion project. This functionality probably will not exist in Subversion 1.0, but it's planned for post-1.0. The problem should be reasonably solvable by recording merge data in 'properties'.]</p> <p>Subversion defines merges the same way CVS does: to merge means to take a set of previously committed changes and apply them, as a patch, to a working copy. This change can then be committed, like any other change. (In Subversion's case, the patch may include changes to directory trees, not just file contents.)</p> <p>As defined thus far, merging is equivalent to hand-editing the working copy into the same state as would result from the patch application. In fact, in CVS there <em>is</em> no difference – it is equivalent to just editing the files, and there is no record of which ancestors these particular changes came from. Unfortunately, this leads to conflicts when users unintentionally merge the same changes again. (Experienced CVS users avoid this problem by using branch- and merge-point tags, but that involves a lot of unwieldy bookkeeping.)</p> <p>In Subversion, merges are remembered by recording <strong class="firstterm">ancestry sets</strong>. A revision's ancestry set is the set of all changes "accounted for" in that revision. By maintaining ancestry sets, and consulting them when doing merges, Subversion can detect when it would apply the same patch twice, and spare users much bookkeeping. Ancestry sets are stored as properties.</p> <p>In the examples below, bear in mind that revision numbers usually refer to changes, rather than the full contents of that revision. For example, "the change A:4" means "the delta that resulted in A:4", not "the full contents of A:4".</p> <p>The simplest ancestor sets are associated with linear histories. For example, here's the history of a file A:</p> <pre> _____ _____ _____ _____ _____ | | | | | | | | | | | A:1 |----->| A:2 |----->| A:3 |----->| A:4 |----->| A:5 | |_____| |_____| |_____| |_____| |_____| </pre> <p>The ancestor set of A:5 is:</p> <pre> { A:1, A:2, A:3, A:4, A:5 } </pre> <p>That is, it includes the change that brought A from nothing to A:1, the change from A:1 to A:2, and so on to A:5. From now on, ranges like this will be represented with a more compact notation:</p> <pre> { A:1-5 } </pre> <p>Now assume there's a branch B based, or "rooted", at A:2. (This postulates an entirely different revision history, of course, and the global revision numbers in the diagrams will change to reflect it.) Here's what the project looks like with the branch:</p> <pre> _____ _____ _____ _____ _____ _____ | | | | | | | | | | | | | A:1 |----->| A:2 |----->| A:4 |----->| A:6 |----->| A:8 |----->| A:9 | |_____| |_____| |_____| |_____| |_____| |_____| \ \ \ _____ _____ _____ \| | | | | | | B:3 |----->| B:5 |----->| B:7 | |_____| |_____| |_____| </pre> <p>If we produce A:9 by merging the B branch back into the trunk</p> <pre> _____ _____ _____ _____ _____ _____ | | | | | | | | | | | | | A:1 |----->| A:2 |----->| A:4 |----->| A:6 |----->| A:8 |---.->| A:9 | |_____| |_____| |_____| |_____| |_____| / |_____| \ | \ | \ _____ _____ _____ / \| | | | | | / | B:3 |----->| B:5 |----->| B:7 |--->-' |_____| |_____| |_____| </pre> <p>then what will A:9's ancestor set be?</p> <pre> { A:1, A:2, A:4, A:6, A:8, A:9, B:3, B:5, B:7} </pre> <p>or more compactly:</p> <pre> { A:1-9, B:3-7 } </pre> <p>(It's all right that each file's ranges seem to include non-changes; this is just a notational convenience, and you can think of the non-changes as either not being included, or being included but being null deltas as far as that file is concerned).</p> <p>All changes along the B line are accounted for (changes B:3-7), and so are all changes along the A line, including both the merge and any non-merge-related edits made before the commit.</p> <p>Although this merge happened to include all the branch changes, that needn't be the case. For example, the next time we merge the B line</p> <pre> _____ _____ _____ _____ _____ _____ _____ | | | | | | | | | | | | | | | A:1 |-->| A:2 |-->| A:4 |-->| A:6 |-->| A:8 |-.->| A:9 |-.->|A:11 | |_____| |_____| |_____| |_____| |_____| | |_____| | |_____| \ / | \ / | \ _____ _____ _____ / _____ | \| | | | | | / | | / | B:3 |-->| B:5 |-->| B:7 |-->|B:10 |->-' |_____| |_____| |_____| |_____| </pre> <p>Subversion will know that A's ancestry set already contains B:3-7, so only the difference between B:7 and B:10 will be applied. A's new ancestry will be</p> <pre> { A:1-11, B:3-10 } </pre> <p>But why limit ourselves to contiguous ranges? An ancestry set is truly a set – it can be any subset of the changes available:</p> <pre> _____ _____ _____ _____ _____ _____ | | | | | | | | | | | | | A:1 |----->| A:2 |----->| A:4 |----->| A:6 |----->| A:8 |--.-->|A:10 | |_____| |_____| |_____| |_____| |_____| / |_____| | / | ______________________.__/ | / | | / | \ __/_ _|__ \ { } { } \ _____ _____ _____ _____ \| | | | | | | | | B:3 |----->| B:5 |----->| B:7 |----->| B:9 |-----> |_____| |_____| |_____| |_____| </pre> <p>In this diagram, the change from B:3-5 and the change from B:7-9 are merged into a working copy whose ancestry set (so far) is { A:1-8 } plus any local changes. After committing, A:10's ancestry set is</p> <pre> { A:1-10, B:5, B:9 } </pre> <p>Clearly, saying "Let's merge branch B into A" is a little ambiguous. It usually means "Merge all the changes accounted for in B's tip into A", but it <em>might</em> mean "Merge the single change that resulted in B's tip into A".</p> <p>Any merge, when viewed in detail, is an application of a particular set of changes – not necessarily adjacent ones – to a working copy. The user-level interface may allow some of these changes to be specified implicitly. For example, many merges involve a single, contiguous range of changes, with one or both ends of the range easily deducible from context (i.e., branch root to branch tip). These inference rules are not specified here, but it should be clear in most contexts how they work.</p> <p>Because each node knows its ancestors, Subversion never merges the same change twice (unless you force it to). For example, if after the above merge, you tell Subversion to merge all B changes into A, Subversion will notice that two of them have already been merged, and so merge only the other two changes, resulting in a final ancestry set of:</p> <pre> { A:1-10, B:3-9 } </pre> <!-- Heh, what about this: B:3 adds line 3, with the text "foo". B:5 deletes line 3. B:7 adds line 3, with the text "foo". B:9 deletes line 3. The user first merges B:5 and B:9 into A. If A had that line, it goes away now, nothing more. Next, user merges B:3 and B:7 into A. The second merge must conflict. I'm not sure we need to care about this, I just thought I'd note how even merges that seem like they ought to be easily composable can still suck. :-) --> <p>This description of merging and ancestry applies to both intra- and inter-repository merges. However, inter-repository merging will probably not be implemented until a future release of Subversion.</p> </div> <!-- model.merging-and-ancestry (h3) --> </div> <!-- model (h2) --> <div class="h2" id="archi" title="#archi"> <h2>Architecture — How Subversion's components work together</h2> <p>Subversion is conceptually divided into a number of separable layers.</p> <p>Assuming that the programmatic interface of each layer is well-defined, it is easy to customize the different parts of the system. Contributors can write new client apps, new network protocols, new server processes, new server features, and new storage back-ends.</p> <p>The following diagram illustrates the "layered" architecture, and where each particular interface lies.</p> <pre> +--------------------+ | commandline or GUI | | client app | +----------+--------------------+----------+ <=== Client interface | Client Library | | | | +----+ | | | | | +-------+--------+ +--------------+--+----------+ <=== Network interface | Working Copy | | Remote | | Local | | Management lib | | Repos Access | | Repos | +----------------+ +--------------+ | Access | | neon | | | +--------------+ | | ^ | | / | | DAV / | | / | | v | | +---------+ | | | | | | | Apache | | | | | | | +---------+ | | | mod_DAV | | | +-------------+ | | | mod_DAV_SVN | | | +----------+-------------+--------------+----------+ <=== Filesystem interface | | | Subversion Filesystem | | | +--------------------------------------------------+ </pre> <div class="h3" id="archi.client" title="#archi.client"> <h3>Client Layer</h3> <p>The Subversion client, which may be either command-line or GUI, draws on three libraries.</p> <p>The working copy library, <tt class="filename">libsvn_wc</tt>, provides an API for managing the client's working copy of a project. This includes operations like renaming or removal of files, patching files, extracting local diffs, and routines for maintaining administrative files in the <tt class="filename">.svn/</tt> directory.</p> <p>The repository_access library, <tt class="filename">libsvn_ra</tt>, provides an API for exchanging information with a Subversion repository. This includes the ability to read files, write new revisions of files, and ask the repository to compare a working copy against its latest revision. Note that there are two implementations of this interface: one designed to talk to a repository over a network, and one designed to work with a repository on local disk. Any number of interface implementations can exist.</p> <p>The client library, <tt class="filename">libsvn_client</tt> provides general client functions such as <tt class="literal">update()</tt> and <tt class="literal">commit()</tt>, which may involve one or both of the other two client libraries. <tt class="filename">libsvn_client</tt> should, in theory, provide an API that allows anyone to write a Subversion client application.</p> <p>For details, see <a href="#client">Client — How the client works</a>.</p> </div> <!-- archi.client (h3) --> <div class="h3" id="archi.network" title="#archi.network"> <h3>Network Layer</h3> <p> The network layer's job is to move the repository API requests over a wire.</p> <p>On the client side, a network library (<tt class="filename">libneon</tt>) translates these requests into a set of HTTP WebDAV/DeltaV requests. The information is sent over TCP/IP to an Apache server. Apache is used for the following reasons:</p> <ul> <li><p>it is time-tested and extremely stable;</p></li> <li><p>it has built-in load-balancing;</p></li> <li><p>it has built-in proxy and firewall support;</p></li> <li><p>it has authentication and encryption features;</p></li> <li><p>it allows client-side caching;</p></li> <li><p>it has an extensible module system</p></li> </ul> <p>Our rationale is that any attempt to write a dedicated "Subversion server" (with a "Subversion protocol") would inevitably end up evolving towards Apache's already-existing feature set. (However, Subversion's layered architecture certainly doesn't <em>prevent</em> anyone from writing a totally new network access implementation.)</p> <p>An Apache module (<tt class="filename">mod_dav_svn</tt>) translates the DAV requests into API calls against a particular repository.</p> <p>For details, see <a href="#protocol">Protocol — How the client and server communicate</a>.</p> </div> <!-- archi.network (h3) --> <div class="h3" id="archi.fs" title="#archi.fs"> <h3>Filesystem Layer</h3> <p>When the requests reach a particular repository, they are interpreted by the <strong class="firstterm">Subversion Filesystem library</strong>, <tt class="filename">libsvn_fs</tt>. The Subversion Filesystem is a custom Unix-like filesystem, with a twist: writes are revisioned and atomic, and no data is ever deleted! This filesystem is currently implemented on top of a normal filesystem, using Berkeley DB files.</p> <p>For a more detailed explanation: see <a href="#server">Server — How the server works</a>.</p> </div> <!-- archi.fs (h3) --> </div> <!-- archi (h2) --> <div class="h2" id="deltas" title="#deltas"> <h2>Deltas — How to describe changes</h2> <p>Subversion uses three kinds of deltas:</p> <ul> <li><p>A <strong><strong class="firstterm">tree delta</strong></strong> describes the difference between two arbitrary directory trees, the way a traditional patch describes the difference between two files. For example, the delta between directories A and B could be applied to A, to produce B.</p> <p>Tree deltas can also carry ancestry information, indicating how the files in one tree are related to files in the other tree. And deltas can describe changes to file meta-information, like permission bits, creation dates, and so on. The repository and working copy use deltas to communicate changes.</p></li> <li><p>A <strong><strong class="firstterm">text delta</strong></strong> describes changes to a string of bytes, such as the contents of a file. It is analogous to traditional patch format, except that it works equally well on binary and text files, and is not invertible (because context and deleted data are not recorded).</p></li> <li><p>A <strong><strong class="firstterm">property delta</strong></strong> describes changes to a list of named properties (see <a href="#model.props">Properties</a>).</p></li> </ul> <p>The term <strong class="firstterm">delta</strong> without qualification generally means a tree delta, unless some other meaning is clear from context.</p> <p>In the examples below, deltas will be described in XML, which happens to be Subversion's (now mostly defunct) import/export patch format. However, note that deltas are an abstract data structure, of which the XML format is merely one representation. Later, we will describe other representations: for example, there is a serialized representation (useful for streaming protocols, among other things), and a db-style representation, used for repository storage. The various representations of a given delta are (in theory, anyway) perfectly isomorphic to one another, since they describe the same underlying structure.</p> <div class="h3" id="deltas.text" title="#deltas.text"> <h3>Text Deltas</h3> <p>A text delta describes the difference between two strings of bytes, the <strong class="firstterm">source</strong> string and the <strong class="firstterm">target</strong> string. Given a source string and a target string, we can compute a text delta; given a source string and a delta, we can reconstruct the target string. However, note that deltas are not invertible: you cannot always reconstruct the source string given the target string and delta.</p> <p>The standard Unix “diff” format is one possible representation for text deltas; however, diffs are not ideal for internal use by a revision control system, for several reasons:</p> <ul> <li><p>Diffs are line-oriented, which makes them human-readable, but sometimes makes them perform poorly on binary files.</p></li> <li><p>Diffs represent a series of replacements, exchanging selected ranges ofthe old text with new text; again, this is easy for humans to read, butit is more expensive to compute and less compact than some alternatives.</p></li> </ul> <p>Instead, Subversion uses the VDelta binary-diffing algorithm, as described in <em class="citetitle">Hunt, J. J., Vo, K.-P., and Tichy, W. F. An empirical study of delta algorithms. Lecture Notes in Computer Science 1167 (July 1996), 49-66.</em> Currently, the output of this algorithm is stored in a custom data format called <strong class="firstterm">svndiff</strong>, invented by Greg Hudson <>, a Subversion developer.</p> <p>The concrete form of a text delta is a well-formed XML element, having the following form:</p> <pre> <text-delta><em class="replaceable">data</em></text-delta> </pre> <p>Here, <em class="replaceable">data</em> is the raw svndiff data, encoded in the MIME Base64 format.</p> </div> <!-- deltas.text (h3) --> <div class="h3" id="deltas.prop" title="#deltas.prop"> <h3>Property Deltas</h3> <p>A property delta describes changes to a property list, of the sort associated with files, directories, and directory entries, and revision numbers (see <a href="#model.props">Properties</a>). A property delta can record creating, deleting, and changing the text of any number of properties.</p> <p>A property delta is an unordered set of name/change pairs. No two pairs within a given property delta have the same name. A pair's name indicates the property affected, and the change indicates what happens to its value. There are two kinds of changes:</p> <dl> <dt>set <em class="replaceable">value</em></dt> <dd><p>Change the value of the named property to the byte string <em class="replaceable">value</em>. If there is no property with the given name, one is added to the property list.</p></dd> <dt>delete</dt> <dd><p>Remove the named property from the property list.</p></dd> </dl> <p>At the moment, the <tt class="literal">set</tt> command can either create or change a property value. However, this simplification means that the server cannot distinguish between a client which believes it is creating a value afresh, and a client which believes it is changing the value of an existing property. It may simplify conflict detection to divide <tt class="literal">set</tt> into two separate <tt class="literal">add</tt> and <tt class="literal">change</tt> operations.</p> <p>In the future, we may add a <tt class="literal">text-delta</tt> change, which specifies a change to an existing property's value as a text delta. This would give us a compact way to describe small changes to large property values.</p> <p>The concrete form of a property delta is a well-formed XML element, having the following form:</p> <pre> <property-delta><em class="replaceable">change</em>…</property-delta> </pre> <p>Each <em class="replaceable">change</em> in a property delta has one of the following forms:</p> <pre> <set name='<em class="replaceable">name</em>'><em class="replaceable">value</em></set> <delete name='<em class="replaceable">name</em>'/> </pre> <p>The <em class="replaceable">name</em> attribute of a <tt class="literal">set</tt> or <tt class="literal">delete</tt> element gives the name of the property to change. The <em class="replaceable">value</em> of a <tt class="literal">set</tt> element gives the new value of the property.</p> <p>If either the property name or the property value contains the characters ‘<tt class="literal">&</tt>’, ‘<tt class="literal"><</tt>’, or ‘<tt class="literal">'</tt>’, they should be replaced with the sequences ‘<tt class="literal">&#38</tt>’, ‘<tt class="literal">&#60</tt>’, or ‘<tt class="literal">&#39</tt>’, respectively.</p> </div> <!-- deltas.prop (h3) --> <div class="h3" id="deltas.tree" title="#deltas.tree"> <h3>Tree Deltas</h3> <p>A tree delta describes changes between two directory trees, the <strong class="firstterm">source tree</strong> and the <strong class="firstterm">target tree</strong>. Tree deltas can describe copies, renames, and deletions of files and directories, changes to file contents, and changes to property lists. A tree delta can also carry information about how the files in the target tree are derived from the files in the source tree, if this information is available.</p> <p>The format for tree deltas described here is easy to compute from a Subversion working directory, and easy to apply to a Subversion repository. Furthermore, the size of a tree delta in this format is independent of the commands used to produce the target tree — it depends only on the degree of difference between the source and target trees.</p> <p>A tree delta is interpreted in the context of three parameters:</p> <ul> <li><p><em class="replaceable">source-root</em>, the name of the directory to which this complete tree delta applies,</p></li> <li><p><em class="replaceable">revision</em>, indicating a particular revision of …</p></li> <li><p><em class="replaceable">source-dir</em>, which is a directory in the source tree that we are currently modifying to yield …</p></li> <li><p>… <strong class="firstterm">target-dir</strong> — the directory we're constructing.</p></li> </ul> <p>When we start interpreting a tree delta, <em class="replaceable">source-root</em>, <em class="replaceable">source-dir</em>, and <em class="replaceable">target-dir</em> are all equal. As we walk the tree delta, <em class="replaceable">target-dir</em> walks the tree we are constructing, and <em class="replaceable">source-dir</em> walks the corresponding portion of the source tree, which we use as the original. <em class="replaceable">Source-root</em> remains constant as we walk the delta; we may use it to choose new source trees.</p> <p>A tree delta is a list of changes of the form</p> <pre> <tree-delta><em class="replaceable">change</em>…</tree-delta> </pre> <p>which describe how to edit the contents of <em class="replaceable">source-dir</em> to yield <em class="replaceable">target-dir</em>. There are three kinds of changes:</p> <dl> <dt><delete name='<em class="replaceable">name</em>'/></dt> <dd><p><em class="replaceable">Source-dir</em> has an entry named <em class="replaceable">name</em>, which is not present in <em class="replaceable">target-dir</em>.</p></dd> <dt><add name='<em class="replaceable">name</em>'><em class="replaceable">content</em></add></dt> <dd><p><em class="replaceable">target-dir</em> has an entry named <em class="replaceable">name</em>, which is not present in <em class="replaceable">source-dir</em>; <em class="replaceable">content</em> describes the file or directory to which the new directory entry refers.</p></dd> <dt><open name='<em class="replaceable">name</em>'><em class="replaceable">content</em></open></dt> <dd><p>Both <em class="replaceable">source-dir</em> and <em class="replaceable">target-dir</em> have an entry named <em class="replaceable">name</em>, which has changed; <em class="replaceable">content</em> describes the new file or directory.</p></dd> </dl> <p>Any entries in <em class="replaceable">source-dir</em> whose names aren't mentioned are assumed to appear unchanged in <em class="replaceable">target-dir</em>. Thus, an empty <tt class="literal">tree-delta</tt> element indicates that <em class="replaceable">target-dir</em> is identical to <em class="replaceable">source-dir</em>.</p> <p>In the change descriptions above, each <em class="replaceable">content</em> takes one of the following forms:</p> <dl> <dt><file <em class="replaceable">ancestor</em>><em class="replaceable">prop-delta</em> <em class="replaceable">text-delta</em></file></dt> <dd><p>The given <em class="replaceable">target-dir</em> entry refers to a file, <em class="replaceable">f</em>. <em class="replaceable">Ancestor</em> indicates which file in the source tree <em class="replaceable">f</em> is derived from, if any. </p> <p><em class="replaceable">Prop-delta</em> is a property delta describing how <em class="replaceable">f</em>'s properties differ from that ancestor; it may be omitted, indicating that the properties are unchanged.</p> <p><em class="replaceable">Text-delta</em> is a text delta describing how to construct <em class="replaceable">f</em> from that ancestor; it may also be omitted, indicating that <em class="replaceable">f</em>'s text is identical to its ancestor's.</p></dd> <dt><file <em class="replaceable">ancestor</em>/></dt> <dd><p>An abbreviation for <tt class="literal"><file <em class="replaceable">ancestor</em>></file></tt> — a fileelement with no property or text delta, thus describing a file identicalto its ancestor.</p></dd> <dt><directory <em class="replaceable">ancestor</em>><em class="replaceable">prop-delta</em> <em class="replaceable">tree-delta</em></directory></dt> <dd><p>The given <em class="replaceable">target-dir</em> entry refers to a subdirectory, <em class="replaceable">sub</em>. <em class="replaceable">Ancestor</em> indicates which directory in the source tree <em class="replaceable">sub</em> is derived from, if any.</p> <p><em class="replaceable">Prop-delta</em> is a property delta describing how <em class="replaceable">sub</em>'sproperties differ from that ancestor; it may be omitted, indicating thatthe properties are unchanged.</p> <p><em class="replaceable">Tree-delta</em> describes how to construct <em class="replaceable">sub</em> from that ancestor; it may be omitted, indicating that the directory is identical to its ancestor. <em class="replaceable">Tree-delta</em> should be interpreted with a new <em class="replaceable">target-dir</em> of <tt class="filename"><em class="replaceable">target-dir</em>/<em class="replaceable">name</em></tt>.</p> <p>Since <em class="replaceable">tree-delta</em> is itself a complete tree delta structure, tree deltas are themselves trees, whose structure is a subgraph of the target tree.</p></dd> <dt><directory <em class="replaceable">ancestor</em>/></dt> <dd><p>An abbreviation for <tt class="literal"><directory <em class="replaceable">ancestor</em>></directory></tt> — a directory element with no property or tree delta, thus describing a directory identical to its ancestor.</p></dd> </dl> <p>The <em class="replaceable">content</em> of a <tt class="literal">add</tt> or <tt class="literal">open</tt> tag may also contain a property delta, describing changes to the properties of that <em>directory entry</em>.</p> <p>In the <tt class="literal">file</tt> and <tt class="literal">directory</tt> elements described above, each <em class="replaceable">ancestor</em> has one of the following forms:</p> <dl> <dt>ancestor='<em class="replaceable">path</em>'</dt> <dd><p>The ancestor of the new or changed file or directory is <tt class="filename"><em class="replaceable">source-root</em>/<em class="replaceable">path</em></tt>, in <em class="replaceable">revision</em>. When this appears as an attribute of a <tt class="literal">file</tt> element, the element's text delta should be applied to <tt class="filename"><em class="replaceable">source-root</em>/<em class="replaceable">path</em></tt>. When this appears as an attribute of a <tt class="literal">directory</tt> element, <tt class="filename"><em class="replaceable">source-root</em>/<em class="replaceable">path</em></tt> should be the new <em class="replaceable">source-dir</em> for interpreting that element's tree delta.</p></dd> <dt>new='true'</dt> <dd><p>This indicates that the file or directory has no ancestor in the source tree. When followed by a <em class="replaceable">text-delta</em>, that delta should be applied to the empty file to yield the new text; when followed by a <em class="replaceable">tree-delta</em>, that delta should be evaluated as if <em class="replaceable">source-dir</em> were an imaginary empty directory.</p></dd> <dt><em class="replaceable">nothing</em></dt> <dd><p>If neither an <tt class="literal">ancestor</tt> nor a <tt class="literal">new</tt> attribute is given, this is an abbreviation for <tt class="literal">ancestor='<em class="replaceable">source-dir</em>/<em class="replaceable">name</em>'</tt>, with the same revision number. This makes the common case — files or directories modified in place — more compact.</p></dd> </dl> <p>If the <em class="replaceable">ancestor</em> spec is not <tt class="literal">new='true'</tt>, it may also contain the text <tt class="literal">revision='<em class="replaceable">rev</em>'</tt>, indicating a new value for <em class="replaceable">revision</em>, in which we should find the ancestor.</p> <p>If a filename or path appearing as a <em class="replaceable">name</em> or <em class="replaceable">path</em> in the description above contains the characters ‘<tt class="literal">&</tt>’, ‘<tt class="literal"><</tt>’, or ‘<tt class="literal">'</tt>’, they should be replaced with the sequences ‘<tt class="literal">&#38;</tt>’, ‘<tt class="literal">&#60;</tt>’, or ‘<tt class="literal">&#39;</tt>’, respectively.</p> <p>Suppose we have the following source tree:</p> <pre> /dir1/file1 file2 dir2/file3 file4 dir3/file5 file6 </pre> <p>If we edit the contents of <tt class="filename">/dir1/file1</tt>, we can describe the effect on the tree with the following tree delta, to be applied to the root:</p> <pre> <tree-delta> <open name='dir1'> <directory> <tree-delta> <open name='file1'> <file><em class="replaceable">text-delta</em></file> </open> </tree-delta> </directory> </open> </tree-delta> </pre> <p>The outer <tt class="literal">tree-delta</tt> element describes the changes made to the root directory. Within the root directory, there are changes in <tt class="filename">dir1</tt>, described by the nested <tt class="literal">tree-delta</tt>. Within <tt class="filename">/dir1</tt>, there are changes in <tt class="filename">file1</tt>, described by the <em class="replaceable">text-delta</em>.</p> <p>If we had edited both <tt class="filename">/dir1/file1</tt> and <tt class="filename">/dir1/file2</tt>, then there would simply be two <tt class="literal">open</tt> elements in the inner <tt class="literal">tree-delta</tt>.</p> <p>As another example, starting from the same source tree, suppose we rename <tt class="filename">/dir1/file1</tt> to <tt class="filename">/dir1/file8</tt>:</p> <pre> <tree-delta> <open name='dir1'> <directory> <tree-delta> <delete name='file1'/> <add name='file8'> <file ancestor='/dir1/file1'/> </add> </tree-delta> </directory> </open> </tree-delta> </pre> <p>As above, the inner <tt class="literal">tdelta</tt> describes how <tt class="filename">/dir1</tt> has changed: the entry for <tt class="filename">/dir1/file1</tt> has disappeared, but there is a new entry, <tt class="filename">/dir1/file8</tt>, which is derived from and textually identical to <tt class="filename">/dir1/file1</tt> in the source directory. This is just an indirect way of describing the rename.</p> <p>Why is it necessary to be so indirect? Consider the delta representing the result of:</p> <ol> <li><p>renaming <tt class="filename">/dir1/file1</tt> to <tt class="filename">/dir1/tmp</tt>,</p></li> <li><p>renaming <tt class="filename">/dir1/file2</tt> to <tt class="filename">/dir1/file1</tt>, and</p></li> <li><p>renaming <tt class="filename">/dir1/tmp</tt> to <tt class="filename">/dir1/file2</tt></p></li> </ol> <p>(in other words, exchanging <tt class="filename">file1</tt> and <tt class="filename">file2</tt>):</p> <pre> <tree-delta> <open name='dir1'> <directory> <tree-delta> <open name='file1'> <file ancestor='/dir1/file2'/> </open> <open name='file2'> <file ancestor='/dir1/file1'/> </open> </tree-delta> </directory> </open> </tree-delta> </pre> <p>The indirectness allows the tree delta to capture an arbitrary rearrangement without resorting to temporary filenames.</p> <p>Another example, starting from the same source tree:</p> <ol> <li><p>rename <tt class="filename">/dir1/dir2</tt> to <tt class="filename">/dir1/dir4</tt>,</p></li> <li><p>rename <tt class="filename">/dir1/dir3</tt> to <tt class="filename">/dir1/dir2</tt>, and</p></li> <li><p>move <tt class="filename">file3</tt> from <em class="replaceable">/dir1/dir4</em> to <em class="replaceable">/dir1/dir2</em>.</p></li> </ol> <p>Note that <tt class="filename">file3</tt>'s path has remained the same, even though the directories around it have changed. Here is the tree delta:</p> <pre> <tree-delta> <open name='dir1'> <directory> <tree-delta> <open name='dir2'> <directory ancestor='/dir1/dir3'> <tree-delta> <add name='file3'> <file ancestor='/dir1/dir2/file3'/> </add> </tree-delta> </directory> </open> <delete name='dir3'/> <add name='dir4'> <directory ancestor='/dir1/dir2'> <tree-delta> <delete name='file3'/> </tree-delta> </directory> </add> </tree-delta> </directory> </open> </tree-delta> </pre> <p>In other words:</p> <ul> <li><p><tt class="filename">/dir1</tt> has changed;</p></li> <li><p>the new directory <tt class="filename">/dir1/dir2</tt> is derived from the old <tt class="filename">/dir1/dir3</tt>, and contains a new entry <tt class="filename">file3</tt>, derived from the old <tt class="filename">/dir1/dir2/file3</tt>;</p></li> <li><p>there is no longer any <tt class="filename">/dir1/dir3</tt>; and</p></li> <li><p>the new directory <tt class="filename">/dir1/dir4</tt> is derived from the old <tt class="filename">/dir1/dir2</tt>, except that its entry for <tt class="filename">file3</tt> is now gone.</p></li> </ul> <p>Some more possible maneuvers, left as exercises for the reader:</p> <ul> <li><p>Delete <tt class="filename">dir2</tt>, and then create a file named <tt class="filename">dir2</tt>.</p></li> <li><p>Rename <tt class="filename">/dir1/dir2</tt> to <tt class="filename">/dir1/dir4</tt>; move <tt class="filename">file2</tt> into <tt class="filename">/dir1/dir4</tt>; and move <tt class="filename">file3</tt> into <em class="replaceable">/dir1/dir3</em>.</p></li> <li><p>Move <tt class="filename">dir2</tt> into <tt class="filename">dir3</tt>, and move <tt class="filename">dir3</tt> into <tt class="filename">/</tt>.</p></li> </ul> </div> <!-- deltas.tree (h3) --> <div class="h3" id="deltas.postfix-text" title="#deltas.postfix-text"> <h3>Postfix Text Deltas</h3> <p>It is sometimes useful to represent a set of changes to a tree without providing text deltas in the middle of the stream. Text deltas are often large and expensive to compute, and tree deltas can be useful without them. For example, one can detect whether two changes might conflict — whether they change the same file, for example — without knowing exactly how the conflicting files changed.</p> <p>For this reason, our XML representation of a tree delta allows the text deltas to come <em>after</em> the </tree-delta> closure. This allows the client to receive early notice of conflicts: during a <tt class="literal">svn commit</tt> command, the client sends a tree-delta to the server, which can check for skeletal conflicts and reject the commit, before the client takes the time to transmit the (possibly large) textual changes. This potentially saves quite a bit of network traffic.</p> <p>In terms of XML, postfix text deltas are split into two parts. The first part appears "in-line" and contains a reference ID. The second part appears after the tree delta is complete. Here's an example:</p> <pre> <tree-delta> <open name="foo.c"> <file> <text-delta-ref id="123"> </file> </open> <add name="bar.c"> <file> <text-delta-ref id="456"> </file> </add> </tree-delta> <text-delta id="123"><em>data</em></text-delta> <text-delta id="456"><em>data</em></text-delta> </pre> </div> <!-- deltas.postfix-text (h3) --> <div class="h3" id="deltas.serializing-via-editor" title="#deltas.serializing-via-editor"> <h3>Serializing Deltas via the "Editor" Interface</h3> <p>The static XML forms above are useful as an import/export format, and as a visualization aid, but we also need a way to express a delta as a <em>series of operations</em>, to implement directory tree diffing and patching. Subversion defines a standard set of such operations in the vtable <tt class="literal">svn_delta_edit_fns_t</tt>, a set of function prototypes which anyone may implement (see <tt class="filename">svn_delta.h</tt>).</p> <p>Each function in an instance of <tt class="literal">svn_delta_editor_t</tt> (colloquially known as an <strong class="firstterm">editor</strong>) implements some distinct subtask of editing a directory tree. In fact, if you compare the editor function prototypes to the XML elements described previously, you'll notice a fairly strict correspondence: there's one function for replacing a directory, another function for replacing a file, one for adding a directory, another for adding a file, a function for deleting, and so on.</p> <p>Although the editor interface was designed around the general idea of making changes to a directory tree, a specific implementation's behavior depends on its role. For example, the versioning filesystem library offers an editor that creates new revisions, while the working copy library offers an editor that updates working copies. And the network layer offers an editor that turns editing calls into wire protocol, which is then converted back into editing calls on the other side! All of these different tasks can share a single interface, because they are all fundamentally about the same thing: expressing and applying differences between directory trees.</p> <p>Like the XML forms, a series of editor calls must follow certain nesting conventions; these conventions are implicit in the interface, in that some of the functions take arguments that can only be obtained from previous calls to other editor functions.</p> <p>Editors can best be understood by watching one work on a real directory tree. For example:</p> <!-- kff todo: fooo working here. --> <p>Suppose that the user has made a number of local changes to her working copy and wants to commit them to the repository. Let's represent her changes with the same tree-delta from a previous example. Notice that she has also made textual modifications to <tt class="filename">file3</tt>; hence the in-line <tt class="literal"><text-delta></tt>:</p> <pre> <tree-delta> <open name='dir1'> <directory> <tree-delta> <open name='dir2'> <directory ancestor='/dir1/dir3'> <tree-delta> <add name='file3'> <file ancestor='/dir1/dir2/file3'> <text-delta><em>data</em></text-delta> </file> </add> </tree-delta> </directory> </open> <delete name='dir3'/> <add name='dir4'> <directory ancestor='/dir1/dir2'> <tree-delta> <delete name='file3'/> </tree-delta> </directory> </add> </tree-delta> </directory> </open> </tree-delta> </pre> <p>So how does the client send this information to the server?</p> <p>In a nutshell: the tree-delta is <em>streamed</em> over the network, as a series of individual commands given in depth-first order.</p> <p>Let's be more specific. The server presents the client with an object of type <tt class="literal">struct svn_delta_edit_fns_t</tt>, colloquially known as an <strong class="firstterm">editor</strong>. An editor is really just table of functions; each function makes a change to a filesystem. Agent A (who has a private filesystem) presents an editor to agent B. Agent B then calls the editor's functions to change A's filesystem. B is said to be <strong class="firstterm">driving</strong> the editor.</p> <p>As Karl Fogel likes to describe the process, if one thinks of the tree-delta as a lion, the editor is a "hoop" that the lion jumps through – each portion of the lion being decomposed through time.</p> <p>B cannot call the functions in any willy-nilly order; there are some logical restrictions. In particular, as B drives the editor, it receives opaque data structures which represent directories and files. It must use and pass these structures, known as <strong class="firstterm">batons</strong>, to make further function calls.</p> <p>As an example, let's watch how the client would transmit the above tree-delta to the repository. (The description below is slightly simplified. For exact interface details, see <tt class="filename">subversion/include/svn_delta.h</tt>.)</p> <p>[Note: in the examples below, and throughout Subversion's code base, you'll see references to 'baton' objects. This is simply a project convention, a name given to structures that define contexts for functions. Many APIs call these structures 'userdata'. In Subversion, we like the term 'baton', because it reminds us of one function “handing off” context to another function.]</p> <ol> <li><p>The repository hands an "editor" to the client.</p></li> <li><p>The client begins by calling <tt class="literal">root_baton = editor->open_root();</tt> The client now has an opaque object, <strong class="firstterm">root_baton</strong>, which represents the root of the repository's filesystem.</p></li> <li><p><tt class="literal">dir1_baton = editor->open_dir("dir1", root_baton);</tt> Notice that <em>root_baton</em> gives the client free license to make any changes it wants in the repository's root directory – until, of course, it calls <tt class="literal">editor->close_dir(root_baton)</tt>. The first change made was a replacement of <tt class="filename">dir1</tt>. In return, the client now has a new opaque data structure that can be used to change <tt class="filename">dir1</tt>.</p></li> <li><p><tt class="literal">dir2_baton = editor->open_dir("dir2", "/dir1/dir3", dir1_baton);</tt> The <em>dir1_baton</em> is now used to open <tt class="filename">dir2</tt> with a directory whose ancestor is <tt class="filename">/dir1/dir3</tt>.</p></li> <li><p><tt class="literal">file_baton = editor->add_file("file3", "/dir1/dir2/file3", dir2_baton);</tt> Edits are now made to <tt class="filename">dir2</tt> (using <em>dir2_baton</em>). In particular, a new file is added to this directory whose ancestor is <tt class="filename">/dir1/dir2/file3</tt>.</p></li> <li><p>Now the text-delta associated with <em>file_baton</em> needs to be transmitted: <tt class="literal">window_handler = editor->apply_textdelta(file_baton);</tt> Text-deltas themselves, for network efficiency, are streamed in "chunks". So instead of receiving a baton object, we now have a routine that is able to receive any number of small "windows" of text-delta data.We won't go into the details of the <tt class="literal">svn_txdelta_*</tt> functions right here; but suffice it to say that these routines are used for sending svndiff data to the <em>window_handler</em> routine.</p></li> <li><p><tt class="literal">editor->close_file(file_baton);</tt> The client is done sending the file's text-delta, so it releases the file baton.</p></li> <li><p><tt class="literal">editor->close_dir(dir2_baton));</tt> The client is done making changes to <tt class="filename">dir2</tt>, so it releases its baton as well.</p></li> <li><p>The client isn't yet finished with <tt class="filename">dir1</tt>, however; it makes two more edits: <tt class="literal">editor->delete_item("dir3", dir1_baton);</tt> <tt class="literal">dir4_baton = editor->add_dir("dir4", "/dir1/dir2", dir1_baton);</tt> <em>(The function's name is <tt class="literal">delete_item</tt> rather than <tt class="literal">delete</tt> to avoid gratuitous incompatibility with C++, where <tt class="literal">delete</tt> is a reserved keyword.)</em></p></li> <li><p>Within the directory <tt class="filename">dir4</tt> (whose ancestry is <tt class="filename">/dir1/dir2</tt>), the client removes a file: <tt class="literal">editor->delete_item("file3", dir4_baton);</tt></p></li> <li><p>The client is now finished with both <tt class="filename">dir4</tt>, as well as its parent <tt class="filename">dir1</tt>: <tt class="literal">editor->close_dir(dir4_baton);</tt> <tt class="literal">editor->close_dir(dir1_baton);</tt></p></li> <li><p>The entire tree-delta is complete. The repository knows this when the root directory is closed: <tt class="literal">editor->close_dir(root_baton);</tt></p></li> </ol> <p>Of course, at any point above, the repository may reject an edit. If this is the case, the client aborts the transmission and the repository hasn't changed a bit. (Thank goodness for transactions!)</p> <p>Note, however, that this "editor interface" works in the other direction as well. When the repository wishes to update a client's working copy, it is the <em>client's</em> reponsibility to give a custom editor-object to the server, and the <em>server</em> is the editor-driver.</p> <p>Here are the main advantages of this interface:</p> <ul> <li><p><em>Consistency</em>. Tree-deltas move across the network, in both directions, using the same interface.</p></li> <li><p><em>Flexibility</em>. Custom editor-implementations can be written to do anything one might want; the editor-driver has no idea what is happening on the other side of the interface. For example, an editor might </p><ul> <li><p>Output XML that matches the tree-delta DTD above;</p></li> <li><p>Output human-readable descriptions of the edits taking place;</p></li> <li><p>Modify a filesystem</p></li> </ul><p> </p></li> </ul> <p>Whatever the case, it's easy to "swap" editors around, and make client and server do new and interesting things.</p> </div> <!-- deltas.serializing-via-editor (h3) --> </div> <!-- deltas (h2) --> <div class="h2" id="client" title="#client"> <h2>Client — How the client works</h2> <p>The Subversion client is built on three libraries. One operates strictly on the working copy and does not talk to the repository. Another talks to the repository but never changes the working copy. The third library uses the first two to provide operations such as <tt class="literal">commit</tt> and <tt class="literal">update</tt> – operations which need to both talk to the repository and change the working copy.</p> <p>The initial client is a Unix-style command-line tool (like standard CVS), but it should be easy to write a GUI client as well, based on the same libraries. The libraries capture the core Subversion functionality, segregating it from user interface concerns.</p> <p>This chapter describes the libraries, and the physical layout of working copies.</p> <div class="h3" id="client.wc" title="#client.wc"> <h3>Working copies and the working copy library</h3> <p>Working copies are client-side directory trees containing both versioned data and Subversion administrative files. The functions in the working copy management library are the only functions in Subversion which operate on these trees.</p> <div class="h4" id="client.wc.layout" title="#client.wc.layout"> <h4>The layout of working copies</h4> <p>This section gives an overview of how working copies are arranged physically, but is not a full specification of working copy layout.</p> <p>As with CVS, Subversion working copies are simply directory trees with special administrative subdirectories, in this case named ".svn" instead of "CVS":</p> <pre> myproj / | \ _____________/ | \______________ / | \ .svn src doc ___/ | \___ /|\ ___/ \___ | | | / | \ | | base ... ... / | \ myproj.texi .svn / | \ ___/ | \___ ____/ | \____ | | | | | | base ... ... .svn foo.c bar.c | ___/ | \___ | | | | | base ... ... myproj.texi ___/ \___ | | foo.c bar.c </pre> <p>Each <tt class="filename">dir/.svn/</tt> directory records the files in <tt class="filename">dir</tt>, their revision numbers and property lists, pristine revisions of all the files (for client-side delta generation), the repository from which <tt class="filename">dir</tt> came, and any local changes (such as uncommitted adds, deletes, and renames) that affect <tt class="filename">dir</tt>.</p> <p>Although it would often be possible to deduce certain information (such as the original repository) by examining parent directories, this is avoided in favor of making each directory be as much a self-contained unit as possible.</p> <p>For example, immediately after a checkout the administrative information for the entire working tree <em>could</em> be stored in one top-level file. But subdirectories instead keep track of their own revision information. This would be necessary anyway once the user starts committing new revisions for particular files, and it also makes it easier for the user to prune a big, complete tree into a small subtree and still have a valid working copy.</p> <p>The <tt class="filename">.svn</tt> subdir contains:</p> <ul> <li><p>A <tt class="filename">format</tt> file, which indicates which version of the working copy adm format this is (so future clients can be backwards compatible easily).</p></li> <li><p>A <tt class="filename">text-base</tt> directory, containing the pristine repository revisions of the files in the corresponding working directory</p></li> <li><p>An <tt class="filename">entries</tt> file, which holds revision numbers and other information for this directory and its files, and records the presence of subdirs. It also contains the repository URLs that each file and directory came from. It may help to think of this file as the functional equivalent of the <tt class="filename">CVS/Entries</tt> file.</p></li> <li><p>A <tt class="filename">props</tt> directory, containing property names and values for each file in the working directory.</p></li> <li><p>A <tt class="filename">prop-base</tt> directory, containing pristine property names and values for each file in the working directory.</p></li> <li><p>A <tt class="filename">dir-props</tt> file, recording properties for this directory.</p></li> <li><p>A <tt class="filename">dir-prop-base</tt> file, recording pristine properties for this directory.</p></li> <li><p>A <tt class="filename">lock</tt> file, whose presence implies that some client is currently operating on the administrative area.</p></li> <li><p>A <tt class="filename">tmp</tt> directory, for holding scratch-work and helping make working copy operations more crash-proof.</p></li> <li><p>A <tt class="filename">log</tt> file. If present, indicates a list of actions that need to be taken to complete a working-copy-operation that is still "in progress".</p></li> </ul> <p>You can read much more about these files in the file <tt class="filename">subversion/libsvn_wc/README</tt>.</p> </div> <!-- client.wc.layout (h4) --> <div class="h4" id="client.wc.library" title="#client.wc.library"> <h4>The working copy management library</h4> <ul> <li><p><strong>Requires:</strong> </p><ul> <li><p>a working copy</p></li> </ul><p> </p></li> <li><p><strong>Provides:</strong> </p><ul> <li><p>ability to manipulate the working copy's versioned data</p></li> <li><p>ability to manipulate the working copy's administrative files</p></li> </ul><p> </p></li> </ul> <p>This library performs "offline" operations on the working copy, and lives in <tt class="filename">subversion/libsvn_wc/</tt>.</p> <p>The API for <em class="replaceable">libsvn_wc</em> is always evolving; please read the header file for a detailed description: <tt class="filename">subversion/include/svn_wc.h</tt>.</p> </div> <!-- client.wc.library (h4) --> </div> <!-- client.wc (h3) --> <div class="h3" id="client.libsvn_ra" title="#client.libsvn_ra"> <h3>The repository access library</h3> <ul> <li><p><strong>Requires:</strong> </p><ul> <li><p>network access to a Subversion server</p></li> </ul><p> </p></li> <li><p><strong>Provides:</strong> </p><ul> <li><p>the ability to interact with a repository</p></li> </ul><p> </p></li> </ul> <p>This library performs operations involving communication with the repository.</p> <p>The interface defined in <tt class="filename">subversion/include/svn_ra.h</tt> provides a uniform interface to both local and remote repository access.</p> <p>Specifically, <em class="replaceable">libsvn_ra_dav</em> will provide this interface and speak to repositories using DAV requests. At some future point, another library <em class="replaceable">libsvn_ra_local</em> will provide the same interface – but will link directly to the filesystem library for accessing local disk repositories.</p> </div> <!-- client.libsvn_ra (h3) --> <div class="h3" id="client.libsvn_client" title="#client.libsvn_client"> <h3>The client operation library</h3> <ul> <li><p><strong>Requires:</strong> </p><ul> <li><p>the working copy management library</p></li> <li><p>a repository access library</p></li> </ul><p> </p></li> <li><p><strong>Provides:</strong> </p><ul> <li><p>all client-side Subversion commands</p></li> </ul><p> </p></li> </ul> <p>These functions correspond to user-level client commands. In theory, any client interface (command-line, GUI, emacs, Python, etc.) should be able to link to <em class="replaceable">libsvn_client</em> and have the ability to act as a full-featured Subversion client.</p> <p>Again, the detailed API can be found in <tt class="filename">subversion/include/svn_client.h</tt>.</p> </div> <!-- client.libsvn_client (h3) --> </div> <!-- client (h2) --> <div class="h2" id="protocol" title="#protocol"> <h2>Protocol — How the client and server communicate</h2> <p>The wire protocol is the connection between the servers, and the client-side <em>Repository Access (RA) API</em>, provided by <tt class="literal">libsvn_ra</tt>. Note that <tt class="literal">libsvn_ra</tt> is in fact only a plugin manager, which delegates the actual task of communicating with a server to one of a selection of back-end modules (the <tt class="literal">libsvn_ra_*</tt> libraries). Therefore, there is not just one Subversion protocol - in fact, at present, there are two:</p> <ul> <li><p>The HTTP/WebDAV/DeltaV based protocol, implemented by the <tt class="literal">mod_dav_svn</tt> Apache 2 server module, and by two independent RA modules, <tt class="literal">libsvn_ra_dav</tt> and <tt class="literal">libsvn_ra_serf</tt>.</p></li> <li><p>The custom-designed protocol built directly upon TCP, implemented by the <tt class="literal">svnserve</tt> server, and the <tt class="literal">libsvn_ra_svn</tt> RA module.</p></li> </ul> <div class="h3" id="protocol.webdav" title="#protocol.webdav"> <h3>The HTTP/WebDAV/DeltaV based protocol</h3> <p>The Subversion client library <tt class="literal">libsvn_ra_dav</tt> uses the <em>Neon</em> library to generate WebDAV DeltaV requests and sends them to a "Subversion-aware" Apache server.</p> <p>This Apache server is running <tt class="literal">mod_dav</tt> and <tt class="literal">mod_dav_svn</tt>, which translates the requests into Subversion filesystem calls.</p> <p>For more info, see <a href="#archi.network">Network Layer</a>.</p> <p>For a detailed description of exactly how Greg Stein <em class="email">gstein@lyra.org</em> is mapping the WebDAV DeltaV spec to Subversion, see his paper: <a href="http://svn.apache.org/repos/asf/subversion/trunk/www/webdav-usage.html">http://svn.apache.org/repos/asf/subversion/trunk/www/webdav-usage.html</a> </p> <p>For more information on WebDAV and the DeltaV extensions, see <a href="http://www.webdav.org">http://www.webdav.org</a> and <a href="http://www.webdav.org/deltav">http://www.webdav.org/deltav</a>. </p> <p>For more information on <em>Neon</em>, see <a href="http://www.webdav.org/neon">http://www.webdav.org/neon</a>.</p> </div> <!-- protocol.webdav (h3) --> <div class="h3" id="protocol.svn" title="#protocol.svn"> <h3>The custom protocol</h3> <p>The client library <tt class="literal">libsvn_ra_svn</tt> and standalone server program <tt class="literal">svnserve</tt> implement a custom protocol over TCP. This protocol is documented at <a href="http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_ra_svn/protocol">http://svn.apache.org/repos/asf/subversion/trunk/subversion/libsvn_ra_svn/protocol</a>.</p> </div> <!-- protocol.svn (h3) --> </div> <!-- protocol (h2) --> <div class="h2" id="server" title="#server"> <h2>Server — How the server works</h2> <p>The term “server” is ambiguous, because it has at least two different meanings: it can refer to a powerful computer which offers services to users on a network, or it can refer to a CPU process designed to receive network requests.</p> <p>In Subversion, however, the <strong class="firstterm">server</strong> is just a set of libraries that implements <strong class="firstterm">repositories</strong> and makes them available to other programs. No networking is required.</p> <p>There are two main libraries: the <strong class="firstterm">Subversion Filesystem</strong> library, and the <strong class="firstterm">Subversion Repository</strong> library.</p> <div class="h3" id="server.fs" title="#server.fs"> <h3>Filesystem</h3> <div class="h4" id="server.fs.overview" title="#server.fs.overview"> <h4>Filesystem Overview</h4> <ul> <li><p><strong>Requires:</strong> </p><ul> <li><p>some writable disk space</p></li> <li><p>(for now) Berkeley DB library</p></li> </ul><p> </p></li> <li><p><strong>Provides:</strong> </p><ul> <li><p>a repository for storing files</p></li> <li><p>concurrent client transactions</p></li> <li><p>enforcement of user & group permissions [someday, not yet]</p></li> </ul><p> </p></li> </ul> <p>This library implements a hierarchical filesystem which supports atomic changes to directory trees, and records a complete history of the changes. In addition to recording changes to file and directory contents, the Subversion Filesystem records changes to file meta-data (see discussion of <strong class="firstterm">properties</strong> in <a href="#model">Model — The versioning model used by Subversion</a>).</p> </div> <!-- server.fs.overview (h4) --> <div class="h4" id="server.fs.api" title="#server.fs.api"> <h4>API</h4> <p> There are two main files that describe the Subversion filesystem.</p> <p>First, read the section below (<a href="#server.fs.struct">Repository Structure</a>) for a general overview of how the filesystem works.</p> <p>Once you've done this, read Jim Blandy's own structural overview, which explains how nodes and revisions are organized (among other things) in the filesystem implementation: <tt class="filename">subversion/libsvn_fs/structure</tt>.</p> <p>Finally, read the well-documented API in <tt class="filename">subversion/include/svn_fs.h</tt>.</p> </div> <!-- server.fs.api (h4) --> <div class="h4" id="server.fs.struct" title="#server.fs.struct"> <h4>Repository Structure</h4> <div class="h5" id="server.fs.struct.schema"> <h5>Schema</h5> <p> To begin, please be sure that you're already casually familiar with Subversion's ideas of files, directories, and revision histories. If not, see <a href="#model">Model — The versioning model used by Subversion</a>. We can now offer precise, technical descriptions of the terms introduced there.</p> <!-- This is taken from jimb's very first Subversion spec! --> <pre> A <strong class="firstterm">text string</strong> is a string of Unicode characters which is canonically decomposed and ordered, according to the rules described in the Unicode standard. A <strong class="firstterm">string of bytes</strong> is what you'd expect. A <strong class="firstterm">property list</strong> is an unordered list of properties. A <strong class="firstterm">property</strong> is a pair <tt class="literal">(<em class="replaceable">name</em>, <em class="replaceable">value</em>)</tt>, where <em class="replaceable">name</em> is a text string, and <em class="replaceable">value</em> is a string of bytes. No two properties in a property list have the same name. A <strong class="firstterm">file</strong> is a property list and a string of bytes. A <strong class="firstterm">node</strong> is either a file or a directory. (We define a directory below.) Nodes are distinguished unions — you can always tell whether a node is a file or a directory. A <strong class="firstterm">node table</strong> is an array mapping some set of positive integers, called <strong class="firstterm">node numbers</strong>, onto <strong class="firstterm">nodes</strong>. If a node table maps some number <em class="replaceable">i</em> to some node <em class="replaceable">n</em>, then <em class="replaceable">i</em> is a <strong class="firstterm">valid node number</strong> in that table, and <strong class="firstterm">node</strong> <em class="replaceable">i</em>is <em class="replaceable">n</em>. Otherwise, <em class="replaceable">i</em> is an <strong class="firstterm">invalid node number</strong> in that table. A <strong class="firstterm">directory entry</strong> is a triple <tt class="literal">(<em class="replaceable">name</em>, <em class="replaceable">props</em>, <em class="replaceable">node</em>)</tt>, where <em class="replaceable">name</em> is a text string, <em class="replaceable">props</em> is a property list, and <em class="replaceable">node</em> is a node number. A <strong class="firstterm">directory</strong> is an unordered list of directory entries, and a property list. A <strong class="firstterm">revision</strong> is a node number and a property list. A <strong class="firstterm">history</strong> is an array of revisions, indexed by a contiguous range of non-negative integers containing 0. A <strong class="firstterm">repository</strong> consists of node table and a history. </pre> <!-- Some definitions: we say that a node @var{n} is a @dfn{direct child} of a directory @var{d} iff @var{d} contains a directory entry whose node number is @var{n}. A node @var{n} is a @dfn{child} of a directory @var{d} iff @var{n} is a direct child of @var{d}, or if there exists some directory @var{e} which is a direct child of @var{d}, and @var{n} is a child of @var{e}. Given this definition of ``direct child'' and ``child,'' the obvious definitions of ``direct parent'' and ``parent'' hold. In these restrictions, let @var{r} be any repository. When we refer, implicitly or explicitly, to a node table without further clarification, we mean @var{r}'s node table. Thus, if we refer to ``a valid node number'' without specifying the node table in which it is valid, we mean ``a valid node number in @var{r}'s node table''. Similarly for @var{r}'s history. --> <p>Now that we've explained the form of the data, we make some restrictions on that form.</p> <p><strong>Every revision has a root directory.</strong> Every revision's node number is a valid node number, and the node it refers to is always a directory. We call this the revision's <strong class="firstterm">root directory</strong>.</p> <p><strong>Revision 0 always contains an empty root directory.</strong> This baseline makes it easy to check out whole projects from the repository.</p> <p><strong>Directories contain only valid links.</strong> Every directory entry's <em class="replaceable">node</em> is a valid node number.</p> <p><strong>Directory entries can be identified by name.</strong> For any directory <em class="replaceable">d</em>, every directory entry in <em class="replaceable">d</em> has a distinct name.</p> <p><strong>There are no cycles of directories.</strong> No node is its own child.</p> <p><strong>Directories can have more than one parent.</strong> The Unix file system does not allow more than one hard link to a directory, but Subversion does allow the analogous situation. Thus, the directories in a Subversion repository form a directed acyclic graph (<strong class="firstterm">DAG</strong>), not a tree. However, it would be distracting and unhelpful to replace the familiar term “directory tree” with the unfamiliar term “directory DAG”, so we still call it a “directory tree” here.</p> <p><strong>There are no dead nodes.</strong> Every node is a child of some revision's root directory.</p> <!-- </jimb> --> </div> <!-- server.fs.struct.schema (h5) --> <div class="h5" id="server.fs.struct.bubble-up"> <h5>Bubble-Up Method</h5> <p>This section provides a conversational explanation of how the repository actually stores and revisions file trees. It's not critical knowledge for a programmer using the Subversion Filesystem API, but most people probably still want to know what's going on “under the hood” of the repository.</p> <p>Suppose we have a new project, at revision 1, looking like this (using CVS syntax):</p> <pre> prompt$ svn checkout myproj U myproj/ U myproj/B U myproj/A U myproj/A/fish U myproj/A/fish/tuna prompt$ </pre> <p>Only the file <tt class="filename">tuna</tt> is a regular file, everything else in myproj is a directory.</p> <p>Let's see what this looks like as an abstract data structure in the repository, and how that structure works in various operations (such as update, commit, and branch).</p> <p>In the diagrams that follow, lines represent parent-to-child connections in a directory hierarchy. Boxes are "nodes". A node is either a file or a directory – a letter in the upper left indicates which kind. A file node has a byte-string for its content, whereas directory nodes have a list of dir_entries, each pointing to another node.</p> <p>Parent-child links go both ways (i.e., a child knows who all its parents are), but a node's name is stored only in its parent, because a node with multiple parents may have different names in different parents.</p> <p>At the top of the repository is an array of revision numbers, stretching off to infinity. Since the project is at revision 1, only index 1 points to anything; it points to the root node of revision 1 of the project:</p> <pre> ( myproj's revision array ) ______________________________________________________ |___1_______2________3________4________5_________6_____... | | ___|_____ |D | | | | A | /* Two dir_entries, `A' and `B'. */ | \ | | B \ | |__/___\__| / \ | \ | \ ___|___ ___\____ |D | |D | | | | | | | | fish | /* One dir_entry, `fish'. */ |_______| |___\____| \ \ ___\____ |D | | | | tuna | /* One dir_entry, `tuna'. */ |___\____| \ \ ___\____ |F | | | | | /* (Contents of tuna not shown.) */ |________| </pre> <p>What happens when we modify <tt class="filename">tuna</tt> and commit? First, we make a new <tt class="filename">tuna</tt> node, containing the latest text. The new node is not connected to anything yet, it's just hanging out there in space:</p> <pre> ________ |F | | | | | |________| </pre> <p>Next, we create a <em>new</em> revision of its parent directory:</p> <pre> ________ |D | | | | tuna | |___\____| \ \ ___\____ |F | | | | | |________| </pre> <p>We continue up the line, creating a new revision of the next parent directory:</p> <pre> ________ |D | | | | fish | |___\____| \ \ ___\____ |D | | | | tuna | |___\____| \ \ ___\____ |F | | | | | |________| </pre> <p>Now it gets more tricky: we need to create a new revision of the root directory. This new root directory needs an entry to point to the “new” directory A, but directory B hasn't changed at all. Therefore, our new root directory also has an entry that still points to the <em>old</em> directory B node!</p> <pre> ______________________________________________________ |___1_______2________3________4________5_________6_____... | | ___|_____ ________ |D | |D | | | | | | A | | A | | \ | | \ | | B \ | | B \ | |__/___\__| |__/___\_| / \ / \ | ___\_____________/ \ | / \ \ ___|__/ ___\____ ___\____ |D | |D | |D | | | | | | | | | | fish | | fish | |_______| |___\____| |___\____| \ \ \ \ ___\____ ___\____ |D | |D | | | | | | tuna | | tuna | |___\____| |___\____| \ \ \ \ ___\____ ___\____ |F | |F | | | | | | | | | |________| |________| </pre> <p>Finally, after all our new nodes are written, we finish the “bubble up” process by linking this new tree to the next available revision in the history array. In this case, the new tree becomes revision 2 in the repository.</p> <pre> ______________________________________________________ |___1_______2________3________4________5_________6_____... | \ | \__________ ___|_____ __\_____ |D | |D | | | | | | A | | A | | \ | | \ | | B \ | | B \ | |__/___\__| |__/___\_| / \ / \ | ___\_____________/ \ | / \ \ ___|__/ ___\____ ___\____ |D | |D | |D | | | | | | | | | | fish | | fish | |_______| |___\____| |___\____| \ \ \ \ ___\____ ___\____ |D | |D | | | | | | tuna | | tuna | |___\____| |___\____| \ \ \ \ ___\____ ___\____ |F | |F | | | | | | | | | |________| |________| </pre> <p>Generalizing on this example, you can now see that each “revision” in the repository history represents a root node of a unique tree (and an atomic commit to the whole filesystem.) There are many trees in the repository, and many of them share nodes.</p> <p>Many nice behaviors come from this model:</p> <ol> <li><p><strong>Easy reads.</strong> If a filesystem reader wants to locate revision <em class="replaceable">X</em> of file <tt class="filename">foo.c</tt>, it need only traverse the repository's history, locate revision <em class="replaceable">X</em>'s root node, then walk down the tree to <tt class="filename">foo.c</tt>.</p></li> <li><p><strong>Writers don't interfere with readers.</strong> Writers can continue to create new nodes, bubbling their way up to the top, and concurrent readers cannot see the work in progress. The new tree only becomes visible to readers after the writer makes its final “link” to the repository's history.</p></li> <li><p><strong>File structure is versioned.</strong> Unlike CVS, the very structure of each tree is being saved from revision to revision. File and directory renames, additions, and deletions are part of the repository's history.</p></li> </ol> <p>Let's demonstrate the last point by renaming the <tt class="filename">tuna</tt> to <tt class="filename">book</tt>.</p> <p>We start by creating a new parent “fish” directory, except that this parent directory has a different dir_entry, one which points the <em>same</em> old file node, but has a different name:</p> <pre> ______________________________________________________ |___1_______2________3________4________5_________6_____... | \ | \__________ ___|_____ __\_____ |D | |D | | | | | | A | | A | | \ | | \ | | B \ | | B \ | |__/___\__| |__/___\_| / \ / \ | ___\_____________/ \ | / \ \ ___|__/ ___\____ ___\____ |D | |D | |D | | | | | | | | | | fish | | fish | |_______| |___\____| |___\____| \ \ \ \ ___\____ ___\____ ________ |D | |D | |D | | | | | | | | tuna | | tuna | | book | |___\____| |___\____| |_/______| \ \ / \ \ / ___\____ ___\____ / |F | |F | | | | | | | | | |________| |________| </pre> <p>From here, we finish with the bubble-up process. We make new parent directories up to the top, culminating in a new root directory with two dir_entries (one points to the old “B” directory node we've had all along, the other to the new revision of “A”), and finally link the new tree to the history as revision 3:</p> <pre> ______________________________________________________ |___1_______2________3________4________5_________6_____... | \ \_________________ | \__________ \ ___|_____ __\_____ __\_____ |D | |D | |D | | | | | | | | A | | A | | A | | \ | | \ | | \ | | B \ | | B \ | | B \ | |__/___\__| |__/___\_| |__/___\_| / ___________________/_____\_________/ \ | / ___\_____________/ \ \ | / / \ \ \ ___|/_/ ___\____ ___\____ _____\__ |D | |D | |D | |D | | | | | | | | | | | | fish | | fish | | fish | |_______| |___\____| |___\____| |___\____| \ \ \ \ \ \ ___\____ ___\____ ___\____ |D | |D | |D | | | | | | | | tuna | | tuna | | book | |___\____| |___\____| |_/______| \ \ / \ \ / ___\____ ___\____ / |F | |F | | | | | | | | | |________| |________| </pre> <p>For our last example, we'll demonstrate the way “tags” and “branches” are implemented in the repository.</p> <p>In a nutshell, they're one and the same thing. Because nodes are so easily shared, we simply create a <em>new</em> directory entry that points to an existing directory node. It's an extremely cheap way of copying a tree; we call this new entry a <strong class="firstterm">clone</strong>, or more colloquially, a “cheap copy”.</p> <p>Let's go back to our original tree, assuming that we're at revision 6 to begin with:</p> <pre> ______________________________________________________ ...___6_______7________8________9________10_________11_____... | | ___|_____ |D | | | | A | | \ | | B \ | |__/___\__| / \ | \ | \ ___|___ ___\____ |D | |D | | | | | | | | fish | |_______| |___\____| \ \ ___\____ |D | | | | tuna | |___\____| \ \ ___\____ |F | | | | | |________| </pre> <p>Let's “tag” directory A. To make the clone, we create a new dir_entry <strong>T</strong> in our root, pointing to A's node:</p> <pre> ______________________________________________________ |___6_______7________8________9________10_________11_____... | \ | \ ___|_____ __\______ |D | |D | | | | | | A | | A | | \ | | | | | B \ | | B | T | |__/___\__| |_/__|__|_| / \ / | | | ___\__/ / / | / \ / / ___|__/ ___\__/_ / |D | |D | | | | | | | | fish | |_______| |___\____| \ \ ___\____ |D | | | | tuna | |___\____| \ \ ___\____ |F | | | | | |________| </pre> <p>Now we're all set. In the future, the contents of directories A and B may change quite a lot. However, assuming we never make any changes to directory T, it will <em>always</em> point to a particular pristine revision of directory A at some point in time. Thus, T is a tag.</p> <p>(In theory, we can use some kind of authorization system to prevent anyone from writing to directory T. In practice, a well-laid out repository should encourage “tag directories” to live in one place, so that it's clear to all users that they're not meant to change.)</p> <p>However, if we <em>do</em> decide to allow commits in directory T, and now our repository tree increments to revision 8, then T becomes a branch. Specifically, it's a branch of directory A which shares history with A up to a certain point, and then “broke off” from the main line at revision 8.</p> </div> <!-- server.fs.struct.bubble-up (h5) --> <div class="h5" id="server.fs.struct.diffy-storage"> <h5>Diffy Storage</h5> <p>You may have been thinking, “Gee, this bubble up method seems nice, but it sure wastes a lot of space. Every commit to the repository creates an entire line of new directory nodes!”</p> <p>Like many other revision control systems, Subversion stores changes as differences. It doesn't make complete copies of nodes; instead, it stores the <em>latest</em> revision as a full text, and previous revisions as a succession of reverse diffs (the word "diff" is used loosely here – for files, it means vdeltas, for directories, it means a format that expresses changes to directories).</p> </div> <!-- server.fs.struct.diffy-storage (h5) --> </div> <!-- server.fs.struct (h4) --> <div class="h4" id="server.fs.implementation" title="#server.fs.implementation"> <h4>Implementation</h4> <p>For the initial release of Subversion,</p> <ul> <li><p>The filesystem will be implemented as a library on Unix.</p></li> <li><p>The filesystem's data will probably be stored in a collection of .db files, using the Berkeley Database library. (In the future, of course, contributors are free modify the Subversion filesystem to operate with more powerful SQL database.) (For more information, see <a href="http://www.sleepycat.com">http://www.sleepycat.com</a>.)</p></li> </ul> </div> <!-- server.fs.implementation (h4) --> </div> <!-- server.fs (h3) --> <div class="h3" id="server.libsvn_repos" title="#server.libsvn_repos"> <h3>Repository Library</h3> <!-- Jimb, Karl: Maybe we should turn this into a discussion about how the filesystem will use non-historical properties for internal ACLs, and how people can add "external" ACL systems via historical properties...? --> <p>A Subversion <strong class="firstterm">repository</strong> is a directory that contains a number of components:</p> <ul> <li><p>a versioned filesystem (typically a collection of .db files)</p></li> <li><p>some hook scripts (for executing before or after commits)</p></li> <li><p>a locking area (used by Berkeley DB or other processes)</p></li> <li><p>a configuration area (for changing global behaviors)</p></li> </ul> <p>The Subversion filesystem is just that: a filesystem. But it's also useful to provide an API that acts at the level of the repository. The repository library (<tt class="filename">libsvn_repos</tt>) does this.</p> <p>In particular, it wraps a few <tt class="filename">libsvn_fs</tt> routines, such as those for beginning and ending commits, so that hook-scripts can run. A pre-commit-hook script might check for a valid log message, and a post-commit-hook script might send an email to a mailing list.</p> <p>Additionally, the repository library provides convenience routines for examining and manipulating the filesystem. For example, a routine to generate a tree-delta by comparing two revisions, routines for constructing new transactions, routines for querying log messages, and routines for exporting and importing filesystem data.</p> </div> <!-- server.libsvn_repos (h3) --> </div> <!-- server (h2) --> <div class="h2" id="license" title="#license"> <h2>License — Copyright</h2> <p>Copyright © 2000-2008 Collab.Net. All rights reserved.</p> <p>This software is licensed as described in the file <tt class="filename">COPYING</tt>, which you should have received as part of this distribution. The terms are also available at <a href="http://subversion.tigris.org/license-1.html">http://subversion.tigris.org/license-1.html</a>. If newer versions of this license are posted there, you may use a newer version instead, at your option.</p> </div> <!-- license (h2) --> </div> </body> </html>