Lint-like program

Richard Uhtenwoldt <rhu@atticus.com>
5 Oct 1998 20:47:27 -0400

          From comp.compilers

Related articles
[Q]: Lint et.al. einulfm@ks.ericsson.se (1998-08-16)
Lint-like program rhu@atticus.com (Richard Uhtenwoldt) (1998-10-05)
| List of all articles for this month |
From: Richard Uhtenwoldt <rhu@atticus.com>
Newsgroups: comp.compilers
Date: 5 Oct 1998 20:47:27 -0400
Organization: Compilers Central
References: 98-08-101
Keywords: tools
cc: Ulf.Magnusson@ks.ericsson.se

This is in response to Ulf Magnusson's query of 2 months ago to
comp.compilers for source code for things like Lint.


I spent several years working on techniques for writing programs that
enforce good style in other programs. To enforce the rules of style I
was interested in, I found I had to translate the program being
checked into CPS and into the lambda calculus. This was a
speculative, pure-research enterprise: I never got as far as figuring
out whether the set of programs whose style I defined to be good could
be compiled into efficient machine code. (I suspected they could,
else I would not have picked that approach; in particular, I took
encouragement from the knowledge that many compilers for Scheme and
for functional-pgmming languages used continuation-passing style as a
intermediate representation.)


The most solid indication I got that I was on the right track was that
I could write a program that ensures that the program being checked
performed only what I had defined to be valid operations on the data
structures in its workspace. An OOPL in which access to an object's
instance variables must go through the object's methods affords
another way to do that, but I like my way better because the OOP way
requires that the data structure remain hidden behind a procedural
abstraction: an OO program seems never to manipulate data directly,
but rather seems always to call methods that do the actual
manipulation. I prefer programs that directly manipulate data--
i:=i+1; c:=cons(foo,c); etc.


I would like to resume this line of investigation, but my research
career is on hold for health reasons.


Ulf Magnusson writes:


>I would be most grateful for hints on papers or implementations that are
>possible for me to download for study and comparison.


I did not get as far as a paper or implementation, so to extract
information from me one would have to engage me in an email
correspondence or other dialog. and it is not clear that the kinds of
style rules that interested me would interest Mr Magnusson.


Post a followup to this message

Return to the comp.compilers page.
Search the comp.compilers archives again.