Subsections

Darcs commands

The general format of a darcs command is

% darcs COMMAND OPTIONS ARGUMENTS ...
Here COMMAND is a command such as add or record, which of course may have one or more arguments. Options have the form --option or -o, while arguments vary from command to command. There are many options which are common to a number of different commands, which will be summarized here.

If you wish, you may use any unambiguous beginning of a command name as a shortcut: for darcs record, you could type darcs recor or darcs rec, but not darcs re since that could be confused with darcs replace, darcs revert and darcs remove.

Common options to darcs commands

Help

Every COMMAND accepts --help as an argument, which tells it to provide a bit of help. Among other things, this help always provides an accurate listing of the options available with that command, and is guaranteed never to be out of sync with the version of darcs you actually have installed (unlike this manual, which could be for an entirely different version of darcs).
% darcs COMMAND --help

Verbose

Most commands also accept the --verbose option, which tells darcs to provide additional output. The amount of verbosity varies from command to command.

Repository directory

Another common option is the --repodir option, which allows you to specify the directory of the repository in which to perform the command. This option is used with commands, such as whatsnew, that ordinarily would be performed within a repository directory, and allows you to use those commands without actually being in the repo directory when calling the command. This is useful when running darcs as a pipe, as might be the case when running apply from a mailer.

Ignore file modification times

Darcs optimizes its operations by keeping track of the modification times of your files. This dramatically speeds up commands such as whatsnew and record which would otherwise require reading every file in the repo and comparing it with a reference version. However, there are times when this can cause problems, such as when running a series of darcs commands from a script, in which case often a file will be modified twice in the same second, which can lead to the second modification going unnoticed. The solution to such predicaments is the --ignore-times option, which instructs darcs not to trust the file modification times, but instead to check each file's contents explicitely.

Author

Several commands need to be able to identify you. Conventionally, you provide an email address for this purpose. The easiest way to do this is to define an environment variable EMAIL or DARCS_EMAIL (with the latter overriding the former). You can also override this using the --author flag to any command. Alternatively, you could set your email address on a per-repository basis using the ``defaults'' mechanism for ``ALL'' commands, as described in Appendix B. Or, you could specify the author on a per-repository basis using the _darcs/prefs/author file as described in section 4.1.3.

Patch compression

By default, darcs commands that write patches to disk will compress the patch files. If you don't want this, you can choose the --dont-compress option, which causes darcs not to compress the patch file.

Optional graphical interface

Certain commands may have an optional graphical user interface. If such commands are supported, you can activate the graphical user interface by calling darcs with the --gui flag.


Resolution of conflicts

To resolve conflicts using an external tool, you need to specify a command to use via something like

--external-merge 'opendiff %1 %2 -ancestor %a -merge %o'.
The %1 and %2 are replaced with the two versions to be merged, %a is replaced with the common ancestor of the two version. Most importantly, %o is replaced with the name of the output file that darcs will require to be created holding the merged version. The above example works with the FileMerge.app tool that comes with Apple's developer tools. To use xxdiff, you would use
--external-merge 'xxdiff -m -O -M %o %1 %a %2'.
If you figure out how to use darcs with another merge tool (meld comes to mind), please let me know what flags you used so I can mention it here.

Options apart from darcs commands

Calling darcs with just ``-help'' as an argument gives a brief summary of what commands are available. Calling darcs with the flag ``-version'' tells you the version of darcs you are using. Similarly calling darcs with only ``-commands'' gives a simple list of available commands. This latter arrangement is primarily intended for the use of command-line autocompletion facilities, as are available in bash.

darcs inittree

Initialize a new source tree as a darcs repository.

Generally you will only call inittree once for each project you work on, and calling it is just about the first thing you do. Just make sure you are in the main directory of the project, and inittree will set up all the directories and files darcs needs in order to start keeping track of revisions for your project.

inittree actually follows a very simple procedure. It simply creates the directories _darcs, _darcs/current and _darcs/patches, and then creates an empty file, _darcs/inventory. However, it is strongly recommended that you use darcs inittree to do this, as this procedure may change in a future version of darcs.

darcs add

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --boring
don't skip boring files
  --case-ok
don't refuse to add files differing only in case
-r --recursive
add contents of subdirectories
  --not-recursive
don't add contents of subdirectories
  --repodir DIRECTORY
specify the repository directory in which to run

Add needs to be called whenever you add a new file or directory to your project. Of course, it also needs to be called when you first create the project, to let darcs know which files should be kept track of.

Darcs will refuse to add a file or directory that differs from an existing one only in case. This is because the HFS+ file system used on under MacOS treats such files as being one and the same.

darcs remove

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --repodir DIRECTORY
specify the repository directory in which to run

Remove should be called when you want to remove a file from your project, but don't actually want to delete the file. Otherwise just delete the file or directory, and darcs will notice that it has been removed.


darcs whatsnew

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --summary
summarize changes
  --no-summary
don't summarize changes
-u --unified
output patch in format similar to diff -u
  --ignore-times
don't trust the file modification times
  --look-for-adds
Add any new files or directories in the working dir
  --dont-look-for-adds
Don't add any files or directories automatically
  --boring
don't skip boring files
  --repodir DIRECTORY
specify the repository directory in which to run

Display unrecorded changes in the working directory.

What's-new gives you a view of what changes you've made in your working directory that haven't yet been recorded. The changes are displayed in darcs patch format. Darcs whatsnew will return a non-zero value if there are no changes, which can be useful if you just want to see in a script if anything has been modified. If you want to see some context around your changes, you can use the -u option, to get output similar to the unidiff format.

If you give one or more file or directory names as an argument to whatsnew, darcs whatsnew will output only changes to those files or to files in those directories.

darcs check

  --complete
check the entire repository
  --partial
check patches since latest checkpoint
-v --verbose
give verbose output
  --quiet
suppress informational output
  --standard-verbosity
neither verbose nor quiet output
  --no-test
don't run the test script
  --test
run the test script

Check the repository for consistency.

Check verifies that the patches stored in the repository, when successively applied to an empty tree, properly recreate the stored current tree.

If you have a checkpoint of the repository (as is the case if you got the repo originally using darcs get --partial), by default darcs check will only verify the contents since the most recent checkpoint. You can change this behavior using the --complete flag.

If you like, you can configure your repository to be able to run a test suite of some sort. You can do this by using ``setpref'' to set the ``test'' value to be a command to run. e.g.

% darcs setpref test "sh configure && make && make test"
Or, if you want to define a test specific to one copy of the repository, you could do this by editing the file _darcs/prefs/prefs.

If you just want to check the consistency of your repository without running the test, you can call darcs check with the --no-test option.

darcs record

-m --patch-name PATCHNAME
name of patch
-A --author EMAIL
specify author id
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --no-test
don't run the test script
  --test
run the test script
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
-a --all
answer yes to all patches
  --pipe
Expect to receive input from a pipe
  --interactive
Prompt user interactively
  --ask-deps
ask about dependencies
  --ignore-times
don't trust the file modification times
  --look-for-adds
Add any new files or directories in the working dir
  --dont-look-for-adds
Don't add any files or directories automatically
  --repodir DIRECTORY
specify the repository directory in which to run

Record is used to name a set of changes and record the patch to the repository. If you provide one or more files or directories as additional arguments to record, you will only be prompted to changes in those files or directories. Each patch is given a name, which typically would consist of a brief description of the changes. This name is later used to describe the patch. The name must fit on one line (i.e. cannot have any embedded newlines). If you have more to say, stick it in the log. The patch is also flagged with the author of the change, taken by default from the DARCS_EMAIL environment variable, and if that doesn't exist, from the EMAIL environment variable. The date on which the patch was recorded is also included. Currently there is no provision for keeping track of when a patch enters a given repository. Finally, each changeset should have a full log (which may be empty). This log is for detailed notes which are too lengthy to fit in the name. If you answer that you do want to create a comment file, darcs will open an editor so that you can enter the comment in. The choice of editor proceeds as follows. If one of the $DARCSEDITOR, $VISUAL or $EDITOR environment variables is defined, its value is used (with precedence proceeding in the order listed). If not, ``vi'', ``emacs'', ``emacs -nw'' and ``nano'' are tried in that order.

There is a problem that shows up with at least some versions of vi which keep them from working properly with darcs. An (ugly) solution to this is to tell vi to use the tty as stdin and stdout:

export DARCSEDITOR="vi </dev/tty >/dev/tty"

Each patch may depend on any number of previous patches. If you choose to make your patch depend on a previous patch, that patch is required to be applied before your patch can be applied to a repo. This can be used, for example, if a piece of code requires a function to be defined, which was defined in an earlier patch.

If you want to manually define any dependencies for your patch, you can use the --ask-deps flag, and darcs will ask you for the patch's dependencies.

If you configure darcs to run a test suite, darcs will run this test on the recorded repo to make sure it is valid. Darcs first creates a pristine copy of the source tree (in /tmp), then it runs the test, using its return value to decide if the record is valid. If it is not valid, the record will be aborted. This is a handy way to avoid making stupid mistakes like forgetting to `darcs add' a new file. It also can be tediously slow, so there is an option (--no-test) to skip the test.

darcs get

  --repo-name REPONAME
name of output repository
-t --tag-name TAGNAME
name of version to pull
-m --patch-name PATCHNAME
name of patch
  --partial
get partial repository using checkpoint
  --complete
get a complete copy of the repository
-v --verbose
give verbose output
  --quiet
suppress informational output
  --standard-verbosity
neither verbose nor quiet output

Get is used to get a local copy of a repository. I recommend using the --verbose flag to get, as this command can take a while, and with no feedback, that can be rather boring.

If the remote repo and the current directory are in the same filesystem and that filesystem supports hard links, get will create hard links for the patch files, which means that the additional storage space needed will be minimal. This is very good for your disk usage (and for the speed of running get), so if you want multiple copies of a repository, I strongly recommend first running darcs get to get yourself one copy, and then running darcs get on that copy to make any more you like. The only catch is that the first time you run darcs push or darcs pull from any of these second copies, by default they will access your first copy--which may not be what you want.

If you want to get a specific version of a repository, you have a couple of options. The first is to use the --tag-name option, which allows you to specify a specific version of the repository by its tag, and get precisely that version. The argument is a regular expression, and get will grab the most recent tag which matches that regular expression. The other option is the --patch-name option, which will allow you to get that patch (i.e. the most recent one matching the specified regular expression) and all older patches. Note that when specifying a --patch-name, you may get a version of your code that has never before been seen, if the patches have gotten themselves reordered. If you ever want to be able to precisely reproduce a given version, you need to tag it.

darcs pull

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
-m --patch-name PATCHNAME
name of patch
-t --tag-name TAGNAME
name of version to pull
-a --all
answer yes to all patches
  --interactive
Prompt user interactively
  --external-merge COMMAND
Use external tool to merge conflicts
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
  --ignore-times
don't trust the file modification times
  --no-deps
don't automatically fulfill dependencies

Pull is used to bring changes made in another repo into the current repo (that is, the one that is the current directory). Pull allows you to bring over all or some of the patches that are in that repo but not in the current one. Pull accepts an argument, which is the URL from which to pull, and when called without an argument, pull will use the repository from which you have most recently either pushed or pulled.

You can use an external interactive merge tool to resolve conflicts via the flag --external-merge. For more details see section 5.1.

The --patch-name argument can be used to specify a regexp, which should be of the extended type used by egrep. If this option is used, only patches which match this regexp (along with their dependencies) are considered. Similarly, --tag-name can be used along with a regexp to pull all patches which are in versions with a matching tag.

If you give a --patch-name argument, darcs will silently pull along any other patches upon which the patches which match the patch-name depend. So --patch-name bugfix mean ``pull all the patches with `bugfix' in their name, along with any patches they require.'' If you really only want the patches with `bugfix' in their name, you should use the --no-deps option, which is only useful in combination with --patch-name, and makes darcs only pull in those matching patches which have no dependencies (apart from other matching patches).

darcs push

-v --verbose
give verbose output
  --quiet
suppress informational output
  --standard-verbosity
neither verbose nor quiet output
-m --patch-name PATCHNAME
name of patch
-a --all
answer yes to all patches
  --interactive
Prompt user interactively
  --apply-as USERNAME
apply patch as another user using sudo
  --apply-as-myself
don't use sudo to apply as another user [DEFAULT]
  --repodir DIRECTORY
specify the repository directory in which to run

Push is the opposite of pull. Push allows you to move changes from the current repository into another repository.

For obvious reasons, you can only push to repositories to which you have write access. In addition, you can only push to repos that you access either via the local file system or via ssh. In order to apply via ssh, darcs must also be installed on the remote computer.

If you give the --apply-as flag, darcs will use sudo to apply the changes as a different user. This can be useful if you want to set up a system where several users can modify the same repository, but you don't want to allow them full write access. This isn't secure against skilled malicious attackers, but at least can protect your repository from clumsy, inept or lazy users.

The --patchname argument can be used to specify a regexp, which should be of the extended type used by egrep. If this option is used, only patches which match this regexp (along with their dependencies) are considered for pushing.

darcs send

-v --verbose
give verbose output
  --quiet
suppress informational output
  --standard-verbosity
neither verbose nor quiet output
-m --patch-name PATCHNAME
name of patch
-a --all
answer yes to all patches
  --interactive
Prompt user interactively
-A --author EMAIL
specify author id
-T --to EMAIL
specify destination email
-c --cc EMAIL
specify email address to cc
-o --output FILE
specify output filename
-s --sign
sign the patch with your gpg key
  --sign-as KEYID
sign the patch with a given keyid
  --sign-ssl IDFILE
sign the patch using openssl with a given private key
  --dont-sign
do not sign the patch
  --edit-description
edit the patch bundle description
  --repodir DIRECTORY
specify the repository directory in which to run

Send is used to prepare a bundle of patches to send to another repository. Send accepts the URL of the repository as an argument. When called without an argument, send will use the most recent repository that was either pushed to, pulled from or sent to.

The --output and --to flags determine what darcs does with the patch bundle after creating it. If you provide an --output argument, the patch bundle is saved to that file. If you give one or more --to arguments, the bundle of patches is emailed to those addresses.

If you don't provide either a --output or a --to flag, darcs will look at the contents of the _darcs/prefs/email file in the target repository (if it exists), and send the patch by email to that address. In this case, you may use the --cc option to specify additional recipients without overriding the default repository email address.

If there is no email address associated with the repository, darcs will prompt you for an email address.

The --patchname argument can be used to specify a regexp, which should be of the extended type used by egrep. If this option is used, only patches which match this regexp (along with their dependencies) are considered for sending.

If you want to include a description or explanation along with the bundle of patches, you need to specify the --edit-description flag, which will cause darcs to open up an editor with which you can compose an email to go along with your patches.

darcs apply

-V --verify PUBRING
verify that the patch was signed by a key in PUBRING
  --verify-ssl KEYS
verify using openSSL with authorized keys from file 'KEYS'
  --no-verify
don't verify patch signature
  --reply FROM
send email response
-c --cc EMAIL
specify email address to cc
  --external-merge COMMAND
Use external tool to merge conflicts
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --ignore-times
don't trust the file modification times
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
-a --all
answer yes to all patches
  --interactive
Prompt user interactively
  --no-resolve-conflicts
don't try to resolve conflicts
  --resolve-conflicts
try to resolve conflicts
  --no-test
don't run the test script
  --test
run the test script
  --repodir DIRECTORY
specify the repository directory in which to run

Apply is used to apply a bundle of patches to this repository. Such a bundle may be created using send.

Darcs apply accepts a single argument, which is the name of the patch file to be applied. If you omit this argument, the patch is read from standard input.5.1 This allows you to use apply with a pipe from your email program, for example.

If you specify the --verify PUBRING option, darcs will check that the patch was gpg-signed by a key which is in PUBRING, and will refuse to apply the patch otherwise.

If you give the --reply FROM option to darcs apply, it will send the results of the application to the sender of the patch. This only works if the patch is in the form of an email with its headers intact, so that darcs can actually know the origin of the patch. The reply email will indicate whether or not the patch was successfully applied. The FROM flag is the email address that will be used as the ``from'' address when replying. If the darcs apply is being done automatically, it is important that this address not be the same as the address at which the patch was received in order to avoid automatic email loops.

If you want to also send the apply email to another address (for example, to create something like a ``commits'' mailing list), you can use the --cc option to specify additional recipients.

The --reply feature of apply is intended primarily for two uses. When used by itself, it is handy for when you want to apply patches sent to you by other developers so that they will know when their patch has been applied. For example, in my .muttrc (the config file for my mailer) I have:

macro pager A "<pipe-entry>darcs apply --verbose \
        --reply droundy@abridgegame.org --repodir ~/darcs
which allows me to apply a patch to darcs directly from my mailer, with the originator of that patch being sent a confirmation when the patch is successfully applied. NOTE: For some reason mutt seems to set the umask such that patches created with the above macro are not world-readable. I'm not sure why this is, but use it with care.

You can use an external interactive merge tool to resolve conflicts via the flag --external-merge. For more details see section 5.1.

If you provide the --interactive or --gui flag, darcs will ask you for each change in the patch bundle whether or not you wish to apply that change. The opposite is the --all flag, which can be used to override an interactive or gui which might be set in your ``defaults'' file.

When used in combination with the --verify option, the --reply option allows for a nice pushable repository. When these two options are used together, any patches that don't pass the verify will be forwarded to the FROM address of the --reply option. This allows you to set up a repository so that anyone who is authorized can push to it and have it automatically applied, but if a stranger pushes to it, the patch will be forwarded to you. Please (for your own sake!) be certain that the --reply FROM address is different from the one used to send patches to a pushable repository, since otherwise an unsigned patch will be forwarded to the repository in an infinite loop.

If you use `darcs apply --verify PUBRING --reply' to create a pushable repo by applying patches automatically as they are recieved via email, you will also want to use the --no-resolve-conflicts option, which will keep the local and recorded versions in sync on the repo.

If you specify the --test option, apply will run the test (if a test exists) prior to applying the patch. If the test fails, the patch is not applied. In this case, if the --reply option was used, the results of the test are send in the reply email. You can also specify the --no-test option, which will override the --test option, and prevent the test from being run. This is helpful when setting up a pushable repository, to keep users from running code.

darcs rollback

-m --patch-name PATCHNAME
name of patch
-t --tag-name TAGNAME
name of version to pull
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --repodir DIRECTORY
specify the repository directory in which to run

Rollback is used to undo the effects of a single patch without actually deleting that patch. Instead, it applies the inverse patch as a new patch. Unlike unpull and unrecord (which accomplish a similar goal) rollback is perfectly safe, since it leaves in the repository a record of the patch it is removing. If you decide you didn't want to roll back a patch after all, you probably should use unrecord to undo the rollback, since like rollback, unrecord doesn't affect the working directory.

darcs unrecord

-m --patch-name PATCHNAME
name of patch
-t --tag-name TAGNAME
name of version to pull
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
  --repodir DIRECTORY
specify the repository directory in which to run

Unrecord is used to undo a single recorded patch. It will prompt you for which patch to unrecord, and then will undo that patch. Unrecord actually removes the unrecorded patch from your repository, so there is no way to ``rerecord'' an unrecorded patch5.2. Note that unrecord doesn't affect your working copy of the tree at all, so if you want to completely undo the change, you'll also need to darcs revert, or do an unpull.

If you don't revert after unrecording, then the changes made by the unrecorded patches are left in your working tree. If these patches are actually from another repository, interaction (either pushes or pulls) with that repository may be massively slowed down, as darcs tries to cope with the fact that you appear to have made a large number of changes that conflict with those present on the other repository. So if you really want to undo the result of a pull operation, use unpull! Unrecord is primarily intended for when you record a patch, realize it needs just one more change, but would rather not have a separate patch for just that one change.

darcs unpull

-m --patch-name PATCHNAME
name of patch
-t --tag-name TAGNAME
name of version to pull
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
  --repodir DIRECTORY
specify the repository directory in which to run

Unpull is used to undo a single patch that has been pulled from another repository. It will prompt you for which patch to unpull, and then will undo that patch. Beware that unpull undoes the patch both from the repo records AND from the current working directory, and does NOT check that the patch originated with a pull. In otherwords, you could lose precious code by unpulling! Like unrecord, unpull does not just apply an inverse patch to the repository, it actually deletes the patch from the repository. This makes unpull a particularly dangerous command, as it not only deletes the patch from the repo, but also removes the changes from the working directory. It is equivalent to an unrecord followed by a revert, except that revert can be unreverted...5.3

darcs revert

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --ignore-times
don't trust the file modification times
-a --all
answer yes to all patches
  --interactive
Prompt user interactively
  --repodir DIRECTORY
specify the repository directory in which to run

Revert is used to undo changes make to the local tree which have not yet been recorded. You will be prompted for which changes you wish to undo. The actions of a revert may be reversed using the unrevert command (see section 5.17). However, if you've made changes since the revert your mileage may vary, so please be careful.

You can give revert optional arguments indicating files or directories. If you do so it will only prompt you to revert changes in those files or in files in those directories.


darcs unrevert

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --ignore-times
don't trust the file modification times
  --repodir DIRECTORY
specify the repository directory in which to run

Unrevert is used to undo the results of a revert command.

While this is only guaranteed to work properly if you haven't made any changes since the revert was performed, it makes a best effort to merge the unreversion with any changes you have since made. In fact, unrevert should even work if you've recorded changes since reverting.

darcs dist

-d --dist-name DISTNAME
name of version
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output

Create a distribution tarball.

Dist is a handy tool for implementing a ``make dist'' target in your makefile. It creates a tarball of the recorded edition of your tree. Basically, you will typically use it via a makefile rule such as

dist:
    ./darcs dist --dist-name darcs-`./darcs --version`
darcs dist then simply creates a clean copy of the source tree, which it then tars and gzips. If you use programs such as autoconf or automake, you really should run them on the clean tree before tarring it up and distributing it. You can do this using the pref value ``predist'', which is a shell command that is run prior to tarring up the distribution:
% darcs setpref predist "autoconf && automake"

darcs mv

  --case-ok
don't refuse to add files differing only in case
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --repodir DIRECTORY
specify the repository directory in which to run

Darcs mv needs to be called whenever you want to rename or move a file or directory. Unlike remove, mv actually performs the move itself in your working directory. This is why ``mv'' isn't called ``move'', since it is really almost equivalent to the unix command ``mv''. I could add an equivalent command named ``move'' for those who like vowels.

Darcs mv will by default refuse to rename a file if there already exists a file having the same name apart from case. This is because doing so could create a repository that could not be used on file systems that are case insensitive (such as Apples HFS+). You can override this by with the flag --case-ok.

darcs replace

  --token-chars "[CHARS]"
define token to contain these characters
  --force
proceed with replace even if 'new' token already exists
  --no-force
don't force the replace if it looks scary
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output

Replace allows you to change a specified token whereever it occurs in the specified files. Tokens here are defined by a regexp specifying the characters which are allowed. By default a token corresponds to a C identifier.

The default regexp is [A-Za-z_0-9]), and if one of your tokens contains a `-' or `.', you will then (by default) get the ``filename'' regexp, which is [A-Za-z_0-9\-\.]. If you prefer to choose a different set of characters to define your token (perhaps because you are programming in some other language), you may do so with the --token-chars option. You may prefer to define tokens in terms of delimiting characters instead of allowed characters using a flag such as --token-chars '[^ \n\t]', which would define a token as being white-space delimited.

If you do choose a non-default token definition, I recommend using _darcs/prefs/defaults to always specify the same --token-chars, since your replace patches will be better behaved (in terms of commutation and merges) if they have tokens defined in the same way.

When using darcs replace, the ``new'' token may not already appear in the file--if that is the case, the replace change would not be invertible. This limitation holds both on the working copy of the file and on the already recorded version of the file.

There is a potentially confusing difference, however, when a replace is used to make another replace possible:

% darcs replace newtoken aaack ./foo.c
% darcs replace oldtoken newtoken ./foo.c
% darcs record
will be valid, even if newtoken and oldtoken are both present in the recorded version of foo.c, while the sequence
% [manually edit foo.c replacing newtoken with aaack]
% darcs replace oldtoken newtoken ./foo.c
will fail because ``newtoken'' still exists in the recorded version of foo.c. The reason for the difference is that when recording, a ``replace'' patch always is recorded before any manual changes, which is usually what you want, since often you will introduce new occurences of the ``newtoken'' in your manual changes. In contrast, ``replace'' changes are recorded in the order in which they were made.

darcs tag

-m --patch-name PATCHNAME
name of patch
-A --author EMAIL
specify author id
  --checkpoint
create a checkpoint file
  --pipe
Expect to receive input from a pipe
  --interactive
Prompt user interactively
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output

Tag is used to name a version of the tree. Tag differs from record in that it doesn't record any new changes, and it always depends on all patches residing in the repository when it is tagged. This means that one can later reproduce this version of the repository by calling, for example:

% darcs get --tag-name "darcs 3.14" REPOLOCATION

Each tagged version has a version name. The version is also flagged with the person who tagged it (taken by default from the `DARCS_EMAIL' or `EMAIL' environment variable). The date is also included in the version information.

A tagged version automatically depends on all patches in the repo. This allows you to later reproduce precisely that version. The tag does this by depending on all patches in the repo, except for those which are depended upon by other tags already in the repo. In the common case of a sequential series of tags, this means that the tag depends on all patches since the last tag, plut that tag itself.

darcs changes

-t --tag-name TAGNAME
name of version to pull
-m --patch-name PATCHNAME
name of patch
  --machine-readable
give machine-readable output
  --xml-output
generate XML formatted output
  --summary
summarize changes
  --no-summary
don't summarize changes
  --repo URL
specify the repository URL

Changes gives a human-readable list of changes between versions suitable for use as a changelog.

When given one or more files or directories as an argument, changes lists only those patches which affect those files or the contents of those directories, or of course the directories themselves.

If changes is given one --tag-name or --patch-name flag, it outputs only those changes since that tag or patch. If two such flags are given, the changes between the two versions are output.

darcs diff

-t --tag-name TAGNAME
name of version to pull
-m --patch-name PATCHNAME
name of patch
  --diff-opts OPTIONS
options to pass to diff
-u  
pass -u option to diff
  --repodir DIRECTORY
specify the repository directory in which to run

Diff can be used to create a diff between two versions which are in your repository. Specifying just one version will get you a diff against the version in your working directory. If you give diff no version arguments, it gives you the same information as whatsnew except that the patch is formatted as the output of a diff command.

Diff calls an external ``diff'' command to do the actual work, and passes any unrecognized flags to this diff command. Thus you can call

% darcs diff -t 0.9.8 -t 0.9.10 -- -u
to get a diff in the unified format. Actually, thanks to the wonders of getopt you need the ``--'' shown above before any arguments to diff. You can also specify additional arguments to diff using the --diff-opts flag. The above command would look like this:
% darcs diff --diff-opts -u -t 0.9.8 -t 0.9.10
This may not seem like an improvement, but it really pays off when you want to always give diff the same options. You can do this by adding
% diff diff-opts -udp
to your _darcs/prefs/defaults file.

If you want to view only the differences to one or more files, you can do so with a command such as

% darcs diff foo.c bar.c baz/

FIXME: I should allow the user to specify the external diff command. Currently it is hardwired to "diff".

darcs setpref

Usage example:

% darcs setpref test "echo I am not really testing anything."

Setpref allows you to set a preferences value in a way that will propagate to other repositories. Valid preferences are: test predist boringfile binariesfile. If you just want to set the pref value in your repository only, you can just edit ``_darcs/prefs/prefs''. Changes you make in that file will be preserved.

The ``_darcs/prefs/prefs'' holds the only preferences information that can propagate between repositories via pushes and pulls, and the only way this happens is when the setprefs command is used. Note that although prefs settings are included in patches, they are not fully version controlled. In particular, depending on the order in which a series of merges is perform, you may end up with a different final prefs configuration. In practice I don't expect this to be a problem, as the prefs usually won't be changed very often.

The following values are valid preferences options which can be configured using setpref:

darcs trackdown

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output

Trackdown tries to find the most recent version in the repository which passes a test. When given no options, trackdown tries to find the latest version that passes the test (i.e. the test that is run under 'darcs record'). If you give it a single argument, it is interpereted as a shell command to be run as a test. If you give it two arguments, the first is a shell command that is run only once (e.g. autoconf, perhaps) and the second is the ``test command''.

Trackdown is helpful for locating when something was broken. FIXME: It is still rather primitive. Currently it just goes back over the history in reverse order trying each version. I'd like for it to explore different patch combinations, to try to find the minimum number of patches that you would need to unpull in order to make the test succeed.

FIXME: I also would like to add an interface by which you can tell it which patches it should consider not including. Without such a feature, the following command:

% darcs trackdown 'make && false'
would result in compiling every version in the repository-which is a rather tedious prospect.

Example usage

If you want to find the last version of darcs that had a FIXME note in the file Record.lhs, you could run
% darcs trackdown 'grep FIXME Record.lhs'

To find the latest version that compiles, you can run

% darcs trackdown 'autoconf' './configure && make'

darcs optimize

  --checkpoint
create a checkpoint file
  --compress
create compressed patches
  --dont-compress
don't create compressed patches
  --uncompress
uncompress patches
-t --tag-name TAGNAME
name of version to pull
-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output

Optimize reorganizes your repository data to make it more efficent to access

Optimize will always write out a fresh copy of the inventory that minimizes the amount of inventory that need be downloaded when people pull from the repo.

If you use the --checkpoint flag optimize creates a checkpoint patch for a tag. You can specify the tag with the --tag-name option, or just let darcs choose the most recent tag.

If you give the compress flag, optimize will compress all the patches in the repository. Similarly, if you give the uncompress or dont-compress flag, optimize will decompress all the patches in the repository. You might want to do this on large repositories if you have plenty of disk space but run into memory pressure, since it will allow darcs to open the patch files via mmap.


darcs annotate

-v --verbose
give verbose output
  --standard-verbosity
don't give verbose output
  --summary
summarize changes
  --no-summary
don't summarize changes
-u --unified
output patch in format similar to diff -u
  --human-readable
give human-readable output
  --xml-output
generate XML formatted output
-m --patch-name PATCHNAME
name of patch
-t --tag-name TAGNAME
name of version to pull
  --repodir DIRECTORY
specify the repository directory in which to run

Display useful information about the repository history.

Annotate allows you to extract all sorts of interesting information from your repository.

When called with just a patch name, annotate outputs the patch in darcs format. The --summary, --human-readable or --unified options may be used to modify the format with which it is displayed.

If a directory name is given, annotate will output the contents of that directory. If a patch name is given, the contents of that directory after that patch was applied will be output. If a tag name is given, the contents of that directory in the specified tagged version will be output. FIXME: None of this is actually implemented.

If a file name is given, the contents of that file will be output, along with markup indicating when each line was last (and perhaps next) modified.

darcs repair

-v --verbose
give verbose output
  --quiet
suppress informational output
  --standard-verbosity
neither verbose nor quiet output

Repair attempts to fix corruption that may have entered your repository.

Repair currently will only repair damage to the _darcs/current directory. Fortunately this is just the sort of corruption that is most likely to happen.

Isaac Jones 2004-04-12