An abstract representation of inline Haddock image URLs with diagrams tags, like <<URL#diagram=name&width=100>>.

Display a diagram URL in the format <<URL#diagram=name&key=val&...>>.

Parse things of the form <<URL#diagram=name&key=val&...>>. The URL is optional (the #, however, is required).

Parse a key/value pair of the form &key=val.

Parse a diagram URL or a single character which is not the start of a diagram URL.

Decompose a string into a parsed form with explicitly represented diagram URLs interspersed with other content.

Serialize a parsed comment with diagram URLs back into a String.

Get the names of all diagrams referenced from diagram URLs in the given comment.

Given a series of comments, return a list of their contents, coalescing blocks of adjacent single-line comments into one String. Each string will be paired with the number of the line on which it begins.

A CodeBlock represents a portion of a comment which is a valid code block (set off by > bird tracks). It also caches the list of bindings present in the code block.

Given a String representing a code block, i.e. valid Haskell code with any bird tracks already stripped off, along with its beginning line number (and the name of the file from which it was taken), attempt to parse it, extract the list of bindings present, and construct a CodeBlock value.

Collect the list of names bound in a module.

From a String representing a comment (along with its beginning line number, and the name of the file it came from, for error reporting purposes), extract all the code blocks (consecutive lines beginning with bird tracks), and error messages for code blocks that fail to parse.

Take the contents of a Haskell source file (and the name of the file, for error reporting purposes), and extract all the code blocks, as well as the referenced diagram names.

Given an identifier and a list of CodeBlocks, filter the list of CodeBlocks to the transitive closure of the "depends-on" relation, i.e. only blocks which bind identifiers referenced in blocks ultimately needed by the block which defines the desired identifier.

Given a directory for cached diagrams and a directory for outputting final diagrams, and all the relevant code blocks, compile the diagram referenced by a single URL, returning a new URL updated to point to the location of the generated diagram. Also return a Bool indicating whether the URL changed.

In particular, the diagram will be output to outDir/name.svg, where outDir is the second argument to compileDiagram, and name is the name of the diagram. The updated URL will also refer to outDir/name.svg, under the assumption that outDir will be copied into the Haddock output directory. (For information on how to make this copying happen, see the README: https://github.com/diagrams/diagrams-haddock/blob/master/README.md.) If for some reason you would like this scheme to be more flexible/configurable, feel free to file a feature request.

Compile all the diagrams referenced in an entire module.

Read a file, compile all the referenced diagrams, and update all the diagram URLs to refer to the proper image files. Note, this overwrites the file, so it's recommended to only do this on files that are under version control, so you can compare the two versions and roll back if processHaddockDiagrams does something horrible.

Returns a list of warnings and/or errors.

Version of processHaddockDiagrams that takes options for cpphs.

Pretty-print a parse failure at a particular location.

A simple monad for collecting a list of error messages. There is no facility for failing as such---in this model one simply generates an error message and moves on.

Generate an error message and fail (by returning Nothing).

Run a CollectErrors computation, resulting in a value of type a along with the collection of generated error messages.