|Managing tokens instead of code tcubed@genesis.MCS.COM (1993-08-08)|
|Re: Managing tokens instead of code email@example.com (1993-08-11)|
|From:||tcubed@genesis.MCS.COM (James Hanlon)|
|Summary:||tokenize, then manage|
|Keywords:||lex, parse, performance, question, comment|
|Organization:||MCSNet Subscriber, Chicago, IL|
|Date:||Sun, 8 Aug 1993 18:39:33 GMT|
A colleague bounced an idea off me: code control and versioning
would be conceptually simpler at the token level, as opposed to the
traditional LOC level. Because tokens are text, too, the old
control/versioning tools work just fine--one must merely compile to the
token level, then stop.
Problem is, compilers in industrial practice don't stop.
Can the tokenize action be made separable, and resumable, in commonly
available (e.g. C) compilers?
[McKeeman and Aki wrote a series of articles on an experimental incremental
C compiler they did at DEC. It sped up recompilation by saving preprocessor
tokens, regular tokens, and partial parses. It appeared in the Journal of C
Language Translation (which I publish) last year. -John]
Return to the
Search the comp.compilers archives again.