rationale.txt

 * Possible use of predefined computational roles: 8/10. We cannot use
   predefined computational roles for all our name analysis because
   it's not syntactically possible to distinguish between defining and
   applied occurrences of symbol names in the concrete grammar section
   of the highlighter specification. This is a limitation due to the
   nature of the BNF grammar that autohighlight inherited from Eli, so
   further improvement is not possible without breaking compatibility.
   Additionally, while it is an error to attempt to highlight a
   literal which doesn't occur in the concrete grammar, our current
   AST doesn't allow for this to be done with predefined computational
   roles. Part of the reason for this is that we so far haven't
   learned to do name analysis on language feature that act like
   records (for example, struct.member access in C).

 * Ease of computation: 9/10. We made a decision to combine all the
   files into one tree, which allows us to do computation across all
   three of them easily. We used the same symbol names across the
   scanner, bnf, and autohighlight grammars so that we could simplify
   computations. However, one thing we weren't able to do is group all
   the symbols that we require for a lot of the computations in one
   portion of the tree, and the shape of the AST doesn't reflect the
   shape of the output file, so we have to do lots of custom output
   and using ptg's might be hard. For instance, to form a simple regex
   rule for a symbol for one of the autohighlighter files, we need to
   get the regex to match the symbol from the gla file portion of the
   tree, the color mapping from the autohighlighter file portion of
   the tree, and the surrounding context from the gla portion of the
   tree. Most of this information will need to be computed because it
   cannot be inferred from the AST.

 * Extensibility: 8/10. Capabilities of text editors change over time.
   New text editors may be supported by autohighlighter in the future.
   The section of autohighlighter most likely to change because of
   added functionality is the color specifications, particularly the
   color criteria such as foreground color and font size). We chose
   not to encode these in the AST by writing concrete rules for every
   color criterion (font-weigh, etc), but rather to treat them as
   identifiers, and handle them using name analysis and type analysis
   on known keys.