[9fans] regexp not greedy enough

[9fans] regexp not greedy enough

[Matthias Teege]

as you know trying to use sam for editing. ;-) I have another
problem with a regexp. I comment out some text with s:(.|\n)*:/*\n&\n*/:
This works but now I try to delete all comments. First I try
to select the region with /\/\*[^\/]*. But the regexp stops at
the first '\n'.

Why? How do I get the whole region?
Matthias

Re: [9fans] regexp not greedy enough

[rog]

> Why? How do I get the whole region?

nothing matches a newline except a newline itself.
so instead of:

	\/\*[^\/]*

i think you might want:

	\/\*([^\/]|\n)*

(BTW, it is possible to write a regexp to properly
match C-style comments, but it's quite involved.
minimal matching would make it quite easy, but i guess
there are good reasons not to implement it)

Re: [9fans] regexp not greedy enough

[Russ Cox]

,x;/\*([^*]|\*[^/]|\n)*\*/;d

Re: [9fans] regexp not greedy enough

[rog]

> ,x;/\*([^*]|\*[^/]|\n)*\*/;d

,x;/\*([^*]|\*[^/]|\*?\n)*\*/;d

does marginally better.

BTW what is it with those:

	elogapply: can't happen 49 49 25

messages anyway?

Re: [9fans] regexp not greedy enough

[Scott Schwartz]

| BTW what is it with those:
| 	elogapply: can't happen 49 49 25
| messages anyway?

There are some scary comments in acme/elog.c that explain it.

Re: [9fans] regexp not greedy enough

[rog]

> There are some scary comments in /sys/src/cmd/acme/elog.c that explain it.

they don't seem to explain why we get these messages
now when we didn't before feb 2003.

but... a little searching back through the dump (oh for a faster
worm!) shows that the only reason that the messages
are printed now (and weren't before) is that the check
was not made before, not that the problem wasn't happening.

so i guess the needling is worth it.

Re: [9fans] regexp not greedy enough

[Russ Cox]

> messages anyway?

they can't happen.  something is wrong.
but i haven't tracked down what.

russ

Re: [9fans] regexp not greedy enough

[boyd, rounin]

> But the regexp stops at the first '\n'.
>
> Why? How do I get the whole region?

usually this means that you need to explicitly stick a \n into the char
class or () expression.

using 's' is a bad idea.  'x' is much more powerful and more appropriate but
takes a bit of a quantum leap to get to grips with it.

did some linux driver hacking with sam this afternoon.  that kernel is
awful, but i think i have tamed it.  sam, on the other hand, is a charm.

Re: [9fans] regexp not greedy enough

[rog]

> using 's' is a bad idea.  'x' is much more powerful and more appropriate but
> takes a bit of a quantum leap to get to grips with it.

'x' is not always sufficient. 's' is sometimes necessary.

RE: [9fans] regexp not greedy enough

[Tiit Lankots]

 
> Why? How do I get the whole region?

,x:/\*: .,/\*\//d

RTFM /sys/doc/sam/sam.ps
 

Re: [9fans] regexp not greedy enough

[Russ Cox]

> ,x:/\*: .,/\*\//d

/* printf("debugging\n"); /**/