chicken-doc-admin

chicken-doc-admin is a tool for generating and maintaining Chicken documentation for the chicken-doc egg.

chicken-doc-admin provides facilities to create and modify a Chicken documentation repository for chicken-doc. It includes a command-line tool to convert egg documentation and manual pages from the Chicken wiki for use with chicken-doc. It also provides an API for repository manipulation.

The typical mode of using chicken-doc-admin is to check out a copy of the wiki using Subversion, then run the chicken-doc-admin command against the copy's egg and man directories. Optionally, process eggdocs by running chicken-doc-admin against an egg repository checkout. Repeat when either is updated; only changed files will be reexamined.

Eggs can also include bundled documentation rather than sourcing it from the wiki; chicken-doc-admin can parse these automatically as well. Typically, though, official eggs use the wiki.

Quick start

The default repository location is in the Chicken install tree and may not be writable by your user. If so, either use sudo or set an alternate writable location with:

$ export CHICKEN_DOC_REPOSITORY=/path/to/some/writable/directory

Initialize the repository if this is your first time:

$ chicken-doc-admin -i

Check out the wiki documentation:

$ svn co --username anonymous --password "" \
     https://code.call-cc.org/svn/chicken-eggs/wiki

To import Chicken 5 documentation:

$ chicken-doc-admin -m wiki/man/5
$ chicken-doc-admin -e wiki/eggref/5

To import Chicken 4 documentation, use 4 instead of 5:

$ chicken-doc-admin -m wiki/man/4
$ chicken-doc-admin -e wiki/eggref/4

chicken-doc-admin is aware of the structure of both Chicken 4 and Chicken 5 manual pages and will automatically import them appropriately. Both Chicken 4 and 5 can serve either documentation version.

Locally installed egg documentation

Certain eggs may install wiki documentation during a chicken-install. Although uncommon, this allows unofficial, private or experimental eggs to include documentation without making it public on the wiki. To do this with your own eggs, see Local egg documentation for developers.

Process all locally installed documentation:

$ chicken-doc-admin -H

Process docs just for the inane and pathfinder eggs:

$ chicken-doc-admin -H inane pathfinder

Working example:

hg clone https://bitbucket.org/ursetto/inane
cd inane
chicken-install
cd ..
chicken-doc-admin -H inane

The cd .. is only for illustrative purposes here to show you can process installed docs from any directory. (Since you checked out from source, you could instead have done chicken-doc-admin -E inane from the source directory to process ./inane.wiki.)

Note: If egg documentation is both on the wiki and locally installed, and you process both, then the latest file timestamp wins. Unfortunately, these timestamps reflect not the actual update time but rather the install time on your local system, and are of dubious value in this case.

Optional eggdoc processing

Almost all users can skip this step, because there are only a few remaining eggs that use eggdoc. The Chicken 5 version of this egg drops support for eggdoc.

Download the egg repository and process any eggdocs:

Warning: eggdocs are executable Scheme code, so if you are concerned about untrusted code, then skip this step or do not run it as root. This risk is reasonably low because anonymous uploads are not allowed.

$ svn co --username anonymous --password "" \
     http://code.call-cc.org/svn/chicken-eggs/release/4 eggs
$ chicken-doc-admin -t eggdoc -e eggs

Upgrading

After an upgrade of the egg, if you get a message informing you that the repository version is invalid, you will need to delete and recreate your repository. For example:

$ chicken-doc-admin -D
$ chicken-doc-admin -i
$ chicken-doc-admin -m ...
$ chicken-doc-admin -e ...

Command-line

$ chicken-doc-admin COMMAND

  -l           list repository information

Print some information about the repository, such as its location and version.

  -i           initialize repository non-destructively

Initialize the repository in the current default location or as overridden by CHICKEN_DOC_REPOSITORY. The directory will be created if it does not exist. This command must be run before any documentation can be processed. You should usually initialize an empty directory; initializing a populated directory won't delete anything, and may confuse us.

  -t type      document type

Set the source document type, which may be eggdoc or svnwiki, defaulting to svnwiki. This option can be used before -e, -E, -m, -M.

To process type eggdoc, you must install the eggdoc-svnwiki extension.

  -e dir       process egg directory DIR

Process a copy of egg directory DIR, adding each egg documentation page to a node at toplevel and the identifiers contained in the page to that node's children. Egg names are displayed as they are processed.

For svnwiki document types, directory is typically /path/to/wiki/eggref/4. The node name will be that of the file's basename, so file eggref/4/9p becomes node 9p.
For eggdoc types, directory is typically /path/to/checkout/of/egg/repo, which is scanned for .meta files containing (eggdoc) attributes. The node name is taken from the eggdoc's (name) attribute.

The timestamp of the source file is stored in each node, and files which have not changed since they were last processed are skipped. Use -f to disregard these timestamps and force reprocessing.

  -m dir       process svnwiki manual directory DIR

Process a copy of svnwiki Chicken manual directory DIR (usually /path/to/wiki/man/4). Adds unit documentation to nodes at toplevel (Unit posix -> posix) and Chicken core documentation to nodes under toplevel node chicken (Parameters -> chicken parameters).

A copy of the manual is usually installed in `(chicken-home)` under `doc/manual` -- for example, `/usr/local/share/chicken/doc/manual` -- and you can use this copy instead of checking it out from SVN.

chicken-doc-admin internally maps each manual page to a node path, based on its filename.

  -E file [path...]     process egg file FILE

Process an individual egg file FILE as in -e. The resulting node path is usually determined from context, but you may set it manually with PATH.

FILE may reside anywhere on disk, not just in a repository checkout.

  -M file [path...]     process svnwiki man file FILE

Process an individual svnwiki manual page FILE as in -m. The node name is determined by an internal map of filenames to node paths. Alternatively, you can set it directly with PATH.

  -H               process all installed egg documentation

Process locally installed egg documentation. Most eggs will keep their documentation on the wiki, but it may also be contained within and installed with the egg itself. This is useful for unofficial, private or experimental eggs.

  -H egg1 egg2...  process specified installed egg documentation

Like -H, but only process the specified eggs.

  -T               process all installed target egg documentation
  -T egg1 egg2...  process specified installed target egg documentation

Like -H, but when Chicken is built as a cross-compiler, it will process eggs installed for the target. When a regular compiler instead, this has the same effect as -H.

  -r           regenerate indices

Regenerate documentation indices; at the moment, this is just one index, the node -> path search map. Indexing is done automatically, so there is no need to use this option unless your index is somehow broken.

  -f           force processing (ignore timestamp checks)

The timestamp of the source file is stored in each node, and normally files which have not changed since they were last processed are skipped. Use -f to disregard these timestamps and force reprocessing. This option can be used before -e, -m, -E and -M, which all consider stored timestamps by default.

  -d path      delete node path recursively

Delete node path PATH and everything under it. Useful for removing, for example, an entire egg from the repository.

WARNING: If you do not provide a path, the root path () is used, which will delete the contents of the entire repository and leave a clean repository in its place.

  -D           destroy repository

Recursively deletes the repository base directory. You must use -i to recreate the repository.

Repository

The repository layout produced by chicken-doc-admin's automatic egg and man parser is detailed in the documentation for the chicken-doc egg.

To recap, documentation for each egg and unit is placed in a toplevel node named after that egg or unit, with procedure, macro, etc. identifiers for that unit as the node's children. Chicken core man pages not corresponding to a particular unit are placed individually under the "chicken" toplevel node; for example, Non-standard macros and special forms resides under the path (chicken macros).

As with chicken-doc, you can override the repository location by setting an environment variable:

export CHICKEN_DOC_REPOSITORY=/path/to/repository

This is useful for testing and also if the default location, which is located in (chicken-home), is not writable by you. You can verify the current repository location with chicken-doc-admin -l.

Low-level repository structure

This structure is subject to change.

|-- .chicken-doc-repo    Repository magic file; contains repository info
|-- id.idx               Alist mapping node identifiers to paths
|-- lock                 Lock file to prevent write conflicts
|-- root/                Documentation root node ()
    |-- 9p/              Documentation node (9p)
    |    |- ,meta         (9p) metadata alist (signature "9p egg", type 'egg)
    |    |- ,sxml         (9p) sxml document
    |    |- ,defs         (9p) definitions (index ...) (def ...) (def ...)
    |
    |-- foreign/         Documentation node (foreign)
         |- ,meta         (foreign) metadata alist + sxml document
         |- access/      Documentation node (foreign access)
            |- ,meta      (foreign access) metadata
            |- ,sxml      (foreign access) sxml document
            |- ,defs      (foreign access) definitions,
                             e.g. for (foreign access foreign-code)

Certain characters are %-escaped in filenames, such as / and period.

When the ,sxml document is under 3KB, it is packed into the ,meta file.

,defs nodes are virtual nodes containing definition sexprs extracted from the parent sxml document at parse time, plus an index. They cannot be created, modified or destroyed except via their containing document, but are otherwise accessed as regular nodes. ,defs nodes first appeared in version 0.4.0; prior to that, they were "real" nodes and one directory was created per definition.

Each node contains a timestamp corresponding to that of the source file.

Proper wiki documentation

This section is a work in progress.

The short and skinny

Use svnwiki type tags (procedure, macro) around all identifiers to be documented. See Chicken-specific wiki syntax.
Place related tagged identifiers on consecutive, non-blank lines; the text description below is then used for all of them. eggdocs must use the proper grouping form, described below.
Ensure all identifier signatures can be read with a call to (read). Don't use characters such as | (pipe).
Ensure open and close tags for bold, italic, links, code and identifiers are located on the same line.

svnwiki identifier type tags

Each identifier is assigned a type (such as 'procedure, 'macro) corresponding to svnwiki tags "procedure", "macro" etc. It is also assigned a signature which is taken verbatim from the svnwiki tag content. Finally, the identifier name also comes from the signature; the signature is parsed using (read), and the result may be a symbol (for example, a constant), or a list --- typically, a procedure or macro signature, in which case the first element of the list is used. With rare exception, as when the signature contains an illegal character such as | (pipe), this strategy works well.

If a signature cannot be parsed as above, the definition is discarded. As a special exception, read syntax signatures are used verbatim, so that you can look up read syntax like #u8.

The currently recognized svnwiki tags are:

(define +identifier-tags+
  (list "procedure" "macro" "read" "parameter"
        "record" "string" "class" "method" "constant" "setter"))

Identifier descriptions

When creating an identifier description, the svnwiki parser takes all text from after the occurrence of the identifier tag up until the next section marker, or the next identifier tag. This follows the informal standard for Chicken documentation on the wiki, and is a natural way to separate descriptions logically. Some points to note:

The description stops even if the section marker denotes a nested section. This works well in general, although there are a few eggs which do contain important identifier info in nested sections. To get this information, you have to read the full documentation, or rearrange the structure.
If several identifier signatures occur on consecutive lines, without any blank lines between them, they are considered part of a "group" of identifiers. The following text description then applies to every identifier in the group. Again, this corresponds to common practice on the wiki.
If you look up one identifier in a group, they will all appear, along with the single text description.

Eggdoc notes

Identifier groups in eggdocs use the somewhat verbose form

(definition
  (signatures
   (signature "procedure" "(foo bar)")
   (signature "procedure" "(baz bat)")
   (signature "macro"     "(quux phlox)"))
  (p "This describes foo, baz and quux."))

Separating the definitions like (procedure ...) (procedure ...) instead defines independent, non-grouped procedures. Nesting (procedure ... (procedure ...)) is illegal.

Local egg documentation for developers

To include local egg documentation, create a text file in svnwiki format, call it <eggname>.wiki, and place it in your toplevel source directory.

foo.setup
foo.meta
foo.scm
foo.wiki

Then in your .setup file, add the .wiki file to your install-extension clause:

(install-extension
  'foo
   '("foo.so" "foo.import.so" "foo.wiki")    ;; add "foo.wiki" here
   '((version "0.0.0")))

Note: Egg documentation that is obtained from the wiki and from installed eggs may cause confusing results for users. The latest file timestamp wins out, but the timestamp reflects local install time, not remote change time. Therefore, the active doc may depend on the user's order of execution of svn co on the wiki, chicken-install of the egg, chicken-doc-admin -e, and chicken-doc-admin -H. In this case, the user may want to run chicken-doc-admin -H eggname explicitly instead of processing all eggs.

0.5.0: Support running on Chicken 5. Workaround file locking bug in Chicken 5.0.0. Repo version 4 for r/w invariance; see chicken-doc 0.6.0. Drop eggdoc support for Chicken 5.
0.4.8: Support Chicken 5 man page layout
0.4.7: Summarize errors, warn (don't crash) if svnwiki parser raises exception.
0.4.6: Fix ID cache bug in 0.4.5 with directory recursion
0.4.5: Recurse into directories. Nicer output format (A/M/?) and summary.
0.4.4: Process locally-installed (host or target) egg documentation
0.4.2: Recognize 'Cross development' manpage
0.4.1: Ensure old definitions are removed from index when updating a node
0.4.0: Repository version 3; dedicated ,defs nodes (saves 80% or more space)
0.3.12: Eliminate manual reindexing; pack page ,sxml into ,meta when < 3072 bytes
0.3.11: Read syntax signatures are used verbatim (requires Chicken 4.4.2 or later)
0.3.9: process only new or modified nodes; display statistics
0.3.7: pack definition ,sxml into ,meta when < 3072 bytes; saves 20+MB
0.3.4: shared repository cache
0.3.0: store pages as SXML; improve parse; repo version 2
0.2.2: Improve manpage handling
0.2: Support eggdoc
0.1.1: Initial release