Use of Interlisp CLISP (infix) in FFILEPOS and elsewhere

[masinter]

Originally posted by @rmkaplan in https://github.com/Interlisp/medley/issues/853#issuecomment-1192178357

However, the original code had another oddity about the treatment of seprs that so far I have carried forward.

The characters that are treated as atom-markers are calculated by the function SEPRCASE, which returns the case array for the characters in a particular readtable. But it also has an argument CLFLG which if T indicates that the characters in CLISPCHARS (+ should be put into the equivalence class along with the read-table's breaks and seprs. CLISPCHARS has the characters ^/+=_:'~+~<>@!←↑. Note in particular that his includes the single quote, which at least in the Interlisp readtable is not a break character--it is a read macro.

In EDITCALLERS, DWIMIFYCOMPFLG is passed as the CLFLG argument, and DWIMIFYCOMPFLG appears to be NIL in my full sysout. The effect is that EDITCALLERS won't find 'XXX.

So, should DWIMIFYCOMPFLG be T? Should CLFLG simply be passed as the constant T? Should single quote be added separately to the equivalence class? I'm not sure why the dependence on DWIMIFYCOMPFLG.

(Separately, I see also that a case array has to be passed explicitly to FFILEPOS (and thus the slow case) if the readtable is case-insensitive. I'll fix that.)

@masinter reply The introduction of Common Lisp and its usurpation of ":" for the package delimiter broke the most important use of CLISP to make RECORD-package references from A:B to (fetch B of A) and A:R.B to (fetch (R B) of A)  So none of the source files in the post-Lyric era have CLISP and CLISP-expansion was put on a FLG basis. As for EDITCALLERS use of FFILEPOS I think you want to err on the side of 

[masinter]

... on the side of recall over precision, since the results of FFILEPOS are filtered by searching the S-expression after READing the results.

[rmkaplan]

So, are you saying that it should always think of CLISPCHARS as atom delimeters (i.e. call SEPRCASE in EDITCALLERS with T)?

Separately, I think quote and comma should also be seen as atom delimiteres, maybe anything in the readtable that is a macro.

(It coud also be tightened up: look before the delimiter to see if it is escaped)

On Jul 22, 2022, at 9:16 AM, Larry Masinter @.***> wrote:

... on the side of recall over precision, since the results of FFILEPOS are filtered by searching the S-expression after READing the results.

— Reply to this email directly, view it on GitHub https://github.com/Interlisp/medley/issues/860#issuecomment-1192734294, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQSTUJJCV63U3R4HZBJPK7DVVLCODANCNFSM54MCNEOA. You are receiving this because you were mentioned.

[masinter]

If you look at man EDITCALLERS it allows prefix and suffix. There's a reference to Editcallers in the "MANAGER" lispusers (which I'd like to put into FULL as a friendlier interface to the file package, especially for CL users). FYI using the output of grep

ShellCommand("grep -boiwr editcallers .") 

b: include byte offsets o: only show what matched (you could filter differently) i: case insensitive match (include files like those in CLOS) w: word delimited (simpler than writing a regexp) r: recursive (give a list of file names instead)

[rmkaplan]

It just says that it allows for escapes, which are interpreted as don’t-check-for-a-delimiter. The question here is, when you don’t have escapes, what atom-delimiters should you be checking for?

(Note that the old code (and carried over into current code) was incorrect in the way it handles escapes in the middle of a search pattern: FOO$FIE. It treats this as if it was just FOO$, not requiring that there be a later FIE presumably before a next delimiter. A little sloppy, but maybe the READ and editor-match compensates for that for EDITCALLERS--but not for FINDCALLERS.)

The w option in the grep command doesn’t actually address our particular functionality, wherein we are trying to use an arbitrary read-table to determine what a “word” is. The seprs/breaks in the readtable presumably could be translated into a regexp without too much trouble.

On Jul 22, 2022, at 10:31 AM, Larry Masinter @.***> wrote:

If you look at man EDITCALLERS it allows prefix and suffix. There's a reference to Editcallers in the "MANAGER" lispusers (which I'd like to put into FULL as a friendlier interface to the file package, especially for CL users). FYI using the output of grep

ShellCommand("grep -boiwr editcallers .") b: include byte offsets o: only show what matched (you could filter differently) i: case insensitive match (include files like those in CLOS) w: word delimited (simpler than writing a regexp) r: recursive (give a list of file names instead)

— Reply to this email directly, view it on GitHub https://github.com/Interlisp/medley/issues/860#issuecomment-1192792466, or unsubscribe https://github.com/notifications/unsubscribe-auth/AQSTUJKFL5VVF3JA44WDCJLVVLLGJANCNFSM54MCNEOA. You are receiving this because you were mentioned.

[masinter]

I'm not sure what this issue is about... my temptation is to turn it into a Discussion since there isn't a problem per see.

If the issue is 'reviving CLISP' there is a lot of design work to do, and I'm not quite sure it's in our goals. s