Synonym for more perspicuous types.

We use Int values for hashes because that's what the Hashable package uses. Assuming diagram hashes are uniformly distributed, on a 64-bit system one needs to build on the order of billions of diagrams before the probability of a hash collision exceeds 1/2, and for anything up to tens of millions of diagrams the probability of a collision is under 0.1%. On 32-bit systems those become tens of thousands and thousands, respectively.

Options to control the behavior of buildDiagram. Create one with mkBuildOpts followed by using the provided lenses to override more fields; for example,

  mkBuildOpts SVG zeroV (Options ...)
    & imports .~ ["Foo.Bar", "Baz.Quux"]
    & diaExpr .~ "square 6 # fc green"

Create a BuildOpts record with default options:

• no snippets
• no pragmas
• no imports
• always regenerate
• the diagram expression circle 1
• no postprocessing

Backend-specific options to use.

Source code snippets. Each should be a syntactically valid Haskell module. They will be combined intelligently, i.e. not just pasted together textually but combining pragmas, imports, etc. separately.

Extra LANGUAGE pragmas to use (NoMonomorphismRestriction is automatically enabled.)

Additional module imports (note that Diagrams.Prelude is automatically imported).

A function to decide whether a particular diagram needs to be regenerated. It will be passed a hash of the final assembled source for the diagram (but with the module name set to Main instead of something auto-generated, so that hashing the source will produce consistent results across runs), plus any options, local imports, and other things which could affect the result of rendering. It can return some information (such as a hash of the source) via the x result, which will be passed through to the result of buildDiagram. More importantly, it decides whether the diagram should be built: a result of Just means the diagram should be built; Nothing means it should not. In the case that it should be built, it returns a function for updating the rendering options. This could be used, e.g., to request a filename based on a hash of the source.

Two standard decision functions are provided for convenience: alwaysRegenerate returns no extra information and always decides to regenerate the diagram; hashedRegenerate creates a hash of the diagram source and looks for a file with that name in a given directory.

The diagram expression to interpret. All the given import sand snippets will be in scope, with the given LANGUAGE pragmas enabled. The expression may have either of the types Diagram b or IO (Diagram b).

A function to apply to the interpreted diagram prior to rendering. For example, you might wish to apply pad 1.1 . centerXY. This is preferred over directly modifying the string expression to be interpreted, since it gives better typechecking, and works no matter whether the expression represents a diagram or an IO action.

Convenience function suitable to be given as the final argument to buildDiagram. It implements the simple policy of always rebuilding every diagram.

Convenience function suitable to be given as the final argument to buildDiagram. It works by converting the hash value to a zero-padded hexadecimal string and looking in the specified directory for any file whose base name is equal to the hash. If there is such a file, it specifies that the diagram should not be rebuilt. Otherwise, it specifies that the diagram should be rebuilt, and uses the provided function to update the rendering options based on the generated hash string. (Most likely, one would want to set the requested output file to the hash followed by some extension.)