Program

From Dyna

A Dyna program defines a set of types, and describes how to compute the values of some items from the values of other items.

Form of a Dyna program

A Dyna program is simply a sequence of inference rules and declarations.

• Each rule or declaration should be terminated by a . (a period).
• The order of rules and declarations is not important.
• Whitespace is not important (except insofar as it affects tokenization).
• The rest of the line following an unquoted % symbol is a comment.

We might switch to using C -style comments in order to allow cpp to run smoothly (see preprocessing). However, I'd like to keep % comments if possible. It's nice for a programmer to be able to glance at a file (or code snippet) and see from the style of comments whether they're looking at the Dyna program or the C driver program.

Preprocessing

In a future version, you should be able to use C preprocessor directives such as #include, #define, and #if ... #endif. These are actually handled by running the Dyna program through cpp, the ISO C preprocessor, so they work just as in a C/C program. For example, #include <lists.dyna> searches the same include directories as in a C program (directories may be added with the -I option to dynac).

For conditional compilation, you'll be able to detect the compiler version (e.g., 0.3.12) via the automatically defined macros __DYNAC__, __DYNAC_MINOR__, and __DYNAC_PATCHLEVEL__. These are by analogy with the GNUC macros for gcc and g .

cpp treats single-quoted strings just like double-quoted ones, protecting them from macro expansion and the like. So the only tricky business with running cpp on Dyna programs is protecting whitespace and Dyna comments.

One solution is to hack cpp.

Another solution seems promising at first, but would unfortunately have to be run on #included files too. Namely, convert Dyna comments to old-fashioned /* C */ comments using a simple filter, then run cpp -C -traditional-cpp (which keeps and protects old-fashioned /* C */ comments, as well as whitespace), and finally run a reverse filter to convert the comments back. Note that the filter must really be reversible, i.e., reversing it shouldn't clobber any actual /* C */ comments in the original Dyna program (e.g., in strings). It should suffice for the filter to change % to /*, /* to /-*, /-* to /--*, etc., and add */ to the end of a line that had at least one %. It's okay to make these changes wherever those character sequences appear, including in quoted strings, and it is actually important to do so in #define statements. One could write the filter and reverse filter directly in C, rather than worrying about whether the user has sed.

Might also need to convert // and /* into something that doesn't look like C comments, in case they are declared as operators in the Dyna program!