Perl select function mystery

macli has asked for the wisdom of the Perl Monks concerning the following question:

After reading perldoc -f select, I still can't get my head around the select function. especially this idom:

($nfound,$timeleft) =
        select($rout=$rin, $wout=$win, $eout=$ein,$timeout);
[download]

What is $rout=$rin, $wout=$win, $eout=$ein? and the return $nfound
Are they refering to default input, output, error output?
How select are used to manage multiple listening sockets?
I often see vec and fileno involved with select.

Anyone can exlain it in plain english to me, or maybe my english is not good engough to understand the perldoc :(

Comment on Perl select function mystery Download Code

Replies are listed 'Best First'.
Re: Perl select function mystery by Joost (Canon) on Mar 27, 2007 at 23:33 UTC
The $rout = $rin stuff makes a copy of the read bits in $rin and passes that to select so that $rout is modified but $rin remains the same. $rout then contains a bitmask of the available file handles (selected from the bitmask in ~~$rout~~ $rin). $nfound is the total number of "ready" filehandles. $wout will contain file handles ready for writing and $eout will contain file handles that have errors. All this is done via bit masks. The bit number corresponds with the fileno of the handle. I.e. bit 0 (value 1) is for fileno 0, bit 1 (value 2) is for fileno 1 etc. At least that how I assume it works. Personally, I find IO::Select much clearer and easier to use. I suggest you take a look at that. update: fixed mistake - thanks rodion :-) "What should it profit a man, if he should win a flame war, yet lose his cool?"	[reply]
Re: Perl select function mystery by duff (Parson) on Mar 28, 2007 at 01:25 UTC
I wrote a select tutorial a while back. You may find it useful. duff	[reply]
Re: Perl select function mystery by Fletch (Bishop) on Mar 27, 2007 at 23:43 UTC
You may also want to look for documentation on the C / NIX `select(2)` system call (such as your system's manual page), or a book like Stevens Advanced Programming in the UNIX Environment*. Seeing the how the underlying system call is used from C may help clear things up (since it's pretty much the same in Perl, you just have to go through an extra step to get the file descriptor from a filehandle). But seconding the IO::Select recommendation, or possibly consider something like POE that provides an event driven framework and hides some of this nasty lower level interaction from you.	[reply]
Re: Perl select function mystery by rodion (Chaplain) on Mar 28, 2007 at 01:09 UTC
How select are used to manage multiple listening sockets? When you need to wait for data coming in on either of two sockets, with file numbers 5 and 8, for example, you set bits 5 and 8 in $rin, and call select($rout=$rin,undef,undef). The select() will wait for data and return when either file number 5 or file number 8 has data to read. You will then need to know which one, and this is why you need $rout which select() will set for you. If bit 5 in $rout is set, then the socket with file number 5 has something to read. If bit 8 is set, then socket number 8. If both are set, then both sockets have something to read. This allows you to only check a socket when you know it's got something you want, and select will wait without chewing up a lot of CPU cycles. If you are writing code instead of reading it, I also second Joost's suggestion of IO::Select.	[reply]
Re: Perl select function mystery by macli (Beadle) on Mar 28, 2007 at 18:59 UTC
Thank everyone for the clarification, duff's tutorial is quite helpful. I am acutally reading the code of Mail::SpamAssassin::SpamdForkScaling which spamd daemon use for children scaling. There is a method main_server_poll which communicates parent with childrens, it involves vec,fileno, select, timeout, error handling in that method. it is fun and pain to read it :)	[reply]