05:22 <Regenaxer> f8l, jan6, ok but I don't even use tput now. Only ANSI escape sequences

10:46 <Regenaxer> Anyway, released a new version today with some bug fixes

12:15 <beneroth> hi all

12:17 <Regenaxer> Hi beneroth

12:17 <beneroth> Regenaxer, how is your availability? Got some time for a pipe optimization problem? :D

12:23 <tankf33der> Regenaxer: new release is ok

12:27 <beneroth> thanks tankf33der

12:30 <beneroth> I'm currently using the following code, how to improve performance? I failed to come up with the code to keep the pipe open for multiple calls, or to replace the pipe with a fifo file, as they need introduction of a signaling char (e.g. one of the special whitespaces intended for that)

12:30 <beneroth> (de html-escape (Str)

12:30 <beneroth> (pipe

12:30 <beneroth> (when Str

12:30 <beneroth> (wait 1)

12:30 <beneroth> (ht:Prin Str)

12:30 <beneroth> (till NIL T) ) ) )

13:14 <Regenaxer> Sorry, driving

13:15 <Regenaxer> You can keep the pipe open if you don't pass a prg

13:16 <Regenaxer> (setq Fd (pipe ...

13:16 <Regenaxer> Why the 'wait'?

13:20 <Regenaxer> eg. 'mail' in @lib/misc.l opens such a persistent pipe

13:21 <Regenaxer> java/lib.l did that too, but is no longer in the distribution

13:22 <Regenaxer> @lib/android.l does it too, but no 'pipe' function (the pipes are opened by the Java part)

13:42 <Regenaxer> java/lib.l basically did (setq *Java (pipe (exec 'java "-cp" Path ...

13:42 <Regenaxer> then communicated bidirectionally with that process

13:42 <Regenaxer> Thanks tankf33der!

13:42 <Regenaxer> (I'm back home btw ;)

14:23 <beneroth> Regenaxer, no problem

14:23 <beneroth> (wait) was added because many quick calls to (html-escape) produced too fast too many fd's

14:25 <beneroth> Regenaxer, yeah I experimented with keeping the (pipe) open, doing only (echo) or so. So then the (ht:Prin) and 'till calls happen both in the caller process, pipe is just abused as fifo. But I didn't manage to get that working, (ht:Prin) always got stuck

14:26 <beneroth> (eof) can not be used as "delimiter", a explicit delimiter is needed for such a permanent I/O loop I think.

14:27 <beneroth> I have to try some more

14:27 <beneroth> not so simple

14:39 <Regenaxer> Ah, yes, too many FDs

14:54 <beneroth> I believe (not profiled) that the creation/closing of the fd's is the performance bottleneck when doing many (html-escape) calls

14:54 <beneroth> unfortunately, I cannot work in-stream (parsing/correcting input, output to be saved in db)

14:55 <Regenaxer> Yes, I would expect open/close be the bottleneck too.

14:56 <Regenaxer> You cannot keep the 'pipe'd process open?

15:19 <beneroth> yes, that I would like to do

15:20 <beneroth> in the one-shot pipe, the prg is just (ht:Prin Str)

15:21 <Regenaxer> I see, not a real external process, jut a pil-helper

15:21 <beneroth> T

15:22 <Regenaxer> (setq *Fd (pipe (exec "picolisp" "-loop (while (line T) (ht:Prin @))" "-bye")))

15:23 <Regenaxer> Ugly :)

15:23 <Regenaxer> (out *Fd (prin ...)) ... (in *Fd (line T))

15:23 <Regenaxer> Something like that, not tested

15:29 <beneroth> thanks, I will test

15:31 <Regenaxer> Wouldn't it be better to modify the string manually?

15:33 <beneroth> is there a hn: function for that? or you mean, re-implementing ht:Prin in picolisp?

15:33 <Regenaxer> yes

15:33 <Regenaxer> eg (pack (replace (chop S) "<" "&lt;" ">" "&gt;" ...))

15:33 <beneroth> probably

15:34 <Regenaxer> Not sure if faster