* Paul Rubin <875xwgzal9.fsf@nightsong.com> :
Wrote on Wed, 17 Apr 2024 00:56:18 -0700:

> Lawrence D'Oliveiro <ldo@nz.invalid> writes:
>> Was it a matter of timing, then? Perl came along at just the point
>> where the hardware was powerful enough to take the complexities of
>> regular expressions in its stride, so that’s when the whole idea
>> really took off.
>
> Unix had regular expressions because Thompson's QED editor on some
> weird old GE(?) minicomputer had had them. It compiled the regexex
> into machine code, iirc. Perl was sort of Awk on steroids and Awk
> also had regexes. I think regexes per se were never very cpu or
> memory hungry.
>
> Snobol and Spitbol didn't have regexes. They did pattern matching by
> brute force backtracking. By that era though, computers had much more
> memory than they did when Lisp 2 was happening.

Languages in the next decade didn't seem to mind not having regex either
REXX, ICON, REBOL (this one is much later)

Jeff Barnett <jbb@notatt.com> wrote or quoted:
>Sir Walter Raleigh was the author of "The Lie"
>Did the Queen know that Raleigh wrote "The Lie"?
>might produce
>Did the Queen know TRUE?

This here thing might have some connection to what folks call
"Extensionality": "To find the value of an expression,
the only thing one needs to know about a subexpression is its
(the subexpression's) value."

Now, we can say with a high degree of certainty that natural
language ain't got no truck with this "Extensionality" business,
at least not the way ol' man McCarthy was demonstratin' it.

The very idea that logical statements are a good fit for
modellin' natural language utterances, and that these bad
boys have some kind of "truth value" as their "meaning"?
Hoo-wee, that's a real doozy, let me tell you!

Paul Rubin <no.email@nospam.invalid> writes:

> Lawrence D'Oliveiro <ldo@nz.invalid> writes:
>> Was it a matter of timing, then? Perl came along at just the point where
>> the hardware was powerful enough to take the complexities of regular
>> expressions in its stride, so that’s when the whole idea really took off.
>
> Unix had regular expressions because Thompson's QED editor on some weird
> old GE(?) minicomputer had had them. It compiled the regexex into
> machine code, iirc. Perl was sort of Awk on steroids and Awk also had
> regexes. I think regexes per se were never very cpu or memory hungry.
>
> Snobol and Spitbol didn't have regexes. They did pattern matching by
> brute force backtracking. By that era though, computers had much more
> memory than they did when Lisp 2 was happening.

SNOBOL's patterns were much more "procedural" than REs. For example,
you could implement Russell's paradox in SNOBOL: a pattern that matches
only those patterns that don't match themselves.

--
Ben.

On Wed, 17 Apr 2024 16:31:58 +0530, Madhu wrote:

> * Paul Rubin <875xwgzal9.fsf@nightsong.com> :
> Wrote on Wed, 17 Apr 2024 00:56:18 -0700:
>
>> Snobol and Spitbol didn't have regexes. They did pattern matching by
>> brute force backtracking. By that era though, computers had much more
>> memory than they did when Lisp 2 was happening.
>
> Languages in the next decade didn't seem to mind not having regex either
> REXX, ICON, REBOL (this one is much later)

None of which are more than museum pieces at this stage. (And you forgot
to mention BASIC.)

Nowadays, say anything about “string processing”, and regexes are
automatically assumed to be part of the mix.

On Wed, 17 Apr 2024 17:44:01 +0100, Ben Bacarisse wrote:

> SNOBOL's patterns were much more "procedural" than REs. For example,
> you could implement Russell's paradox in SNOBOL: a pattern that matches
> only those patterns that don't match themselves.

I would expect a paradox like that to behave the same way it does when
expressed in λ-calculus: get stuck in an endless loop.

Mathematicians seem to be scared of paradoxes. But if you realize that
they are just equivalent to endless loops in a computation, which is
something we deal with all the time in Comp Sci, then they are no longer
so scary.

> Mathematicians seem to be scared of paradoxes. But if you realize that
> they are just equivalent to endless loops in a computation, which is
> something we deal with all the time in Comp Sci, then they are no longer
> so scary.

The problem is that mathematicians don't run their proofs, so such an
inf-loop trips them up into thinking that everything is true.

It's also a problem for computers when we don't want to run our proofs
either (e.g. when we don't want to wait for that proof that the index is
within bounds to finish before we access the array?).

Stefan

Lawrence D'Oliveiro <ldo@nz.invalid> writes:
> Nowadays, say anything about “string processing”, and regexes are
> automatically assumed to be part of the mix.

Haskell has a regex library but people tend to use one of the parser
combinator libraries instead (Parsec or Attoparsec). Parsec reminds me
a bit of SNOBOL patterns though the mechanism is not at all similar.

SNOBOL was a very old fashioned funky language. Every statement
optionally contained up to two different GOTOs. ICON was more modern
looking. I actually used SNOBOL as as a kid, and it was fun. But yes,
museum piece.

I'm not old enough to have been around for when large applications were
written in assembly language, but SPITBOL was one of those applications,
and perusing the source code is very impressive.