Say Perl

Hello.

I want to share with you, my dear friends, a new reference: shootnix.com . Long-held desire to write my little pieces of code that solve the problem more perlish, finally realized in the form Posterous -blozhika.

Some time ago I tried to write small posts in the style of «Perl so powerful ...", in which he proposed to solve the problem whichever way in Perl: time , two . We are pleased to have found this subject interesting not only for me.

Well, what then is continued in a more befitting place? Duc! =)

I did not expect this, but YAML:: Dumper and YAML are soooo sloooow. On the laptop with i7 gen2 (!!!) visibility delay of 1 + second one just use the module.

Well, there is a YAML:: Tiny.

 use HTML:: Entities;
...
my $ fixed = encode_entities ($ raw, "\ x {460} - \ x {2E2E}");

 my $ fixed = encode_entities ($ raw, "\ x {A0} - \ x {40F} \ x {460} - \ x {2E2E}"); 

Over the past week:
- Exactly seven nights spent in trying to make a new version of my photo hosting
- Just 8:00 gone in to make the first working prototype dancerctl .

ps: git add. & & Git commit & & git push instead of svn commit - it's, well, excuse me. I would like on this occasion to pronounce the words that should not know the gentleman.

We look for Perl / C programmer udalenku with extensive experience in the development of frontend / backend heavy-sites.

Attendance million unique visitors per day to the server. Iron works at full capacity.

All the software is written in Perl + MySQL + third-party template. Need to constantly introduce new additions to improve performance, up to rewriting frontend / backend completely. Plus auxiliary software to write Perl.

Just need to configure the server (Free BSD) to the needs of the software. Those issues that will decide the admins, you can redirect them unnecessarily to the full server administration, hosting providers.

ZP on probation 2 months (in conjunction with the old computer programmer) - 45 tys.r ($ 1,500) / month. then 75t.r. ($ 2,500) / month. In the long term may be extended up to 120 tys.r ($ 4,000) for performance. On probation can pay for a week. Payment WebMoney WMZ or at one of the card payment systems.

Schedule a free, 7-8 hours a day. With good performance, maybe less. Other part-time work are not allowed on the side, unnecessarily just to think about our software.

Mandatory condition - a permanent presence in the ICQ and the ability to connect to the Internet no later than 1-2 hours after my call at any time, in case of an emergency, for example, did not rise after soft reboot the server.

You can contact me via PM, but it is better to ICQ 1 7 - 2 two - 0 4

ARGGHhh! I have already copied at least three times the extension of domain architecture, and even rewrote it so that it was correct in terms of architecture to find a simple mistake with the missing parameters in a function call ...

The current version of Padre some hellish nightmare with the clipboard. And terribly short of a custom avtopodstanovki.

And the saddest thing - for opensurs projects now really better to be on github-e. And so - I do not even know what to do Padre. There's sort of like you can even become committers, but this process will require reading the text in the heap they are even terrible wiki, writing letters to someone in general totally inhumane. I have a year like laziness. I'm in this totally not a fan of git-a, but the example of a Dancer-and studs around him very clearly that here for such a project - github - this is what you need.

Poyuzat Padre few days of the trunk. Fall sick, I put 0.92. But the trend is sad - the last really stable version was 0.86.

0.88 I dropped in the first three minutes of use - it may fall if the files were opened, podtseplennye via sshfs, which broke. Or it may not fall. Something tells me here that the normal testing the product at all. In previous versions of full stocks with sliding illuminated and non-working sometimes a quick search.

Ah. Of course, I'll put the GraphicsMagick and use it, but with the orientation of development is clearly something wrong.

Mercurial, on the Download page makes no reference to the source (I'm only 10 minutes later I saw that in the main menu there is a separate link to Source), download the Perl Graphics:: Magick, rather, that much more valuable to read its documentation on search.cpan.org - not. When installing packages from Graphics:: Magick is placed, but to put it in a separate ubuntu, which is not obvious, because the documentation on the site says that it is placed immediately with gm.

Well, that too sadly, the stable version of the year with a tail, is unstable, as amended, and new chips on 7 of my 22-inch screens, but it is clear that unstable there better not to put into production. In general, would be stable releases at least once in three months - it would be much better.

upd: there still is a wonderful documentation for Perl bindings ...

Foreword

December 18, the 24th birthday of Pearl, the third time will be Russian Pearl Workshop Saint Perl .

Participation is free. Registration is open (you can login with their password from the previous events).

We invite speakers, exhibitors and sponsors! business incubator " Ingria "for the provision of facilities.

Saint Perl - the 26th event in a series of YAPC:: Russia , founded in 2007 and covering the territory of seven countries: Rosiii, Ukraine, Belarus, Kazakhstan, Uzbekistan, Bulgaria and Latvia.

1. 
   
2. Perl script for sending SMS
   
3. Why Perl? (Translated into Russian)
   
4. How do I change Perl syntax to your taste
   
5. Back from BlackPerl 2011
   
6. vti about "small gifts" and a very interesting project (in perl, esessno) privytalks.com
   
7. Performance packages in Perl OOP
   
8. (Question) Perl, and database testing
   
9. Mojolicious And Dancer
   
10. Larry Wall's birthday:)
    

1. 
   
2. Login to hh.ru and find on-demand "Perl"
   
3. (!) MAIL.RU - PERL programmer (cool Perl - about - bison)
   

1. 
   
2. AI:: ParticleSwarmOptimization - cm http://goo.gl/JvRQM
   
3. Google:: Plus
   
4. Set:: Functional
   
   

1. 
   
2. Of information . Chic series of articles about compression algorithms already in 6 parts.
   
3. Try to cluster file system GridFS (the one that at MongoDB) + benchmark
   
4. In bitbaket now has Git
   

1. Perl script for sending SMS
2. Why Perl? (Translated into Russian)
3. How do I change Perl syntax to your taste
4. Back from BlackPerl 2011
5. vti about "small gifts" and a very interesting project (in perl, esessno) privytalks.com
6. Performance packages in Perl OOP
7. (Question) Perl, and database testing
8. Mojolicious And Dancer
9. Larry Wall's birthday:)

http://www.slideshare.net/koorchik/maybe-you-do-not-know-that

Frame 34 - real private methods

 my $ name = sub {
    my $ self = shift;
    ...
}

$ Self-> $ name ();
$ Self-> $ name ('viktor'); 

ps: good, but service - slideshare. One only frustrating - to find something really interesting - have to scroll through dozens of presentations. And there is copy-paste does not work - you have to download the presentation locally, open to something local and only then. It was very uncomfortable.

Anatoly Sharifulin translation prepared presentation Jesse Vincent Perl 5.16 and Beyond , which he visited on several recent YAPC.

exiftool '-filename<${shuttercount}.%e' .

On the eve of the first anniversary of last autumn DST in Russia recalled a fakap three years ago (early 2008). I then engaged in repairing the attendance statistics on a regional Login or register, which I supported as a friend.

They had two servers, one on the internet show content to visitors, and the second was located in the office and kept the interface to administer our website. All farm work under FreeBSD 6.1 on very old machines, on the main server was a Pentium 3 and 384 MB of RAM (however many years ago about such a configuration, one could only dream ). Apache 2.2.2 kept a log of visits in the mode of «combined», and at night they all accumulated in the last day was treated full-applets logresolve (converts IP-addresses of visitors to host names) and copied to a separate file resolved.log. Then pick up an office server resulting file over HTTP (for security reasons all the "alternative" methods of file sharing have decided not to use) and parses it awstats'om. Such is the ingenuous shemki. But it worked for exactly two months until mid-December 2006 and ceased.

Apart from a few bugs found a couple of strategic failure. First, the office server sometimes disconnected due to problems with electricity, and the log is not taken in time, and the next day just to be supplemented by, and after two or three days off with logs file has grown to such proportions that even in a compressed form for too long passed would be to http, and taking away the script will simply fall off the timeout. Second, there is established logresolve naturally spoiled logs: read lines from the first 1024 bytes, handles IP-address, and published them in STDOUT, completing the transfer line, and then read the following 1024 bytes, etc. As a result, the rows that contain, for example, long referrerskie URL, were broken into several lines of 1024 bytes each, which may not matter to awstats, but it is extremely difficult to additional processing of the log.

As a result, from December 2006 to the log file grew in size up to six gigabytes. I decided it was best to break it for months and months to feed these parser turns, but first had to deal with the extra line breaks. We quickly sketched on the one-liner sed, which removes extra line breaks, if not in the third field contains the date.

However, despite the fact that both small and large enough to present excerpts resolved.log skriptik working out the relatively normal, the log of real read only the first 170 megabytes of coppers and the script hangs for a long time. I decided that maybe something did not realize because of the small experience of this kind of writing scripts for sed, and rewrote the same algorithm in perl and c - to no avail: after reading the first 170 megabytes, nothing happened.

 my $ b ='';
while (<>){
        chomp;
        if (/ \ [[0-9] + \ / [a-zA-Z] + \ / [0-9:] + \ \ +0300 \] /) {
                print "$ b \ n";
                $ B = $ _;
        } Else {
                $ B. = $ _
        }
}
print "$ b \ n";

Several times I drove the scripts through valgrind with maximum leakage checks, get rid of some of the mistakes in the end read a lot of documentation on freebsd on possible restrictions on the size of the process in memory (although more than eight kilobytes of program memory is not occupied), etc. . In general, for two days, dropped out of life, spending time in debugging ...

Finally found the cause of my anguish: the regular expression defining feature of the recording log, was hard-coded time zone 0300, while the transition to summer time it changed in 0400. Therefore, when the cycle reached in March 2007, several hundred thousand lines until next October just skipped, and because the server was slow, then Ctrl + C and kill-TERM scripts to react immediately, and to create a sense of hovering.

Once I corrected this regexps, everything went like clockwork: fed awstats'u old logs, tied logrotate, taught scripts to collect statistics over scp; parallel crown set anacron , so that the fence and log analysis was performed on a daily basis regardless of whether or not the "office" server in 4 hours and 15 minutes in the morning. And then did criminal

#! / Usr / bin / perl-w
use EV;
use AnyEvent;
use Spread;
use YAML;
use Twiggy:: Server;
use open IO => ': encoding (utf8)';
use strict;

$ | = 1;
sub LOG {
   print STDERR join ('', map {ref ($_)? Dump ($ _): $ _} map {defined $ _? $ _: '_'} @ _), "\ n";
};

my ($ mbox, $ group) = Spread:: connect ({spread_name => '4803 ', private_name =>' psgi '});
die Spread:: error () unless defined $ mbox;

Spread:: join ($ mbox, $ _) for qw (dial connect hangup);

my% conn;
my $ w01 = AE:: io $ mbox, 0, sub {
   my ($ service_type, $ sender, $ groups, $ mess_type, $ endian, $ message) = Spread:: receive ($ mbox);
   foreach my $ k (keys% conn) {
      $ Conn {$ k} -> {writer} -> write ('data:'. $ Message. "\ N \ n");
   }
};

LOG ('startup');
Twiggy:: Server-> new (host => '0 .0.0.0 ', port => '5001') -> register_service (sub {
   my $ env = shift;
   return sub {
      my $ respond = shift;
      my $ writer = $ respond-> ([200, ['Content-Type', 'text / event-stream']]);
      my $ fh = fileno ($ env-> {'psgix.io'});
      $ Writer-> {handle} -> on_error (sub {
         LOG ('Connection close', $ fh);
         delete $ conn {$ fh};
      });
      LOG ('adding conn to pool', $ fh, $ env-> {HTTP_USER_AGENT});
      $ Conn {$ fh} = {env => $ env, writer => $ writer};
   };
});

EV:: loop;

server {
  ...
  location / event {
    proxy_pass http://127.0.0.1:5001;
    proxy_buffering off;
  }

  location / {
    proxy_pass http://127.0.0.1:5000;
  }
}

There has recently been discussed, but the post was removed, and there were interesting commentary that I have not had time to keep, so he decided to raise the issue again.

perl, Dancer, like when you start the application instances skonnektitsya with musklem and further, if possible, use this connection and not make a connection before {} and disconnection in after {}. Here is a descent pridumyvetsya:

 $ Dbh = DBI-> connect ("DBI: mysql: dbname =". $ Dbname. "; Host =". $ Dbhost, $ dbuser, $ dbpass, { 
             PrintError => 0, 
                AutoCommit => 1, 
                mysql_enable_utf8 => 1,
          mysql_auto_reconnect => 1,
        }) 
or return undef; 

At first glance, it works. Should I consider something else?

From a cursory reading of the sources I do not understand how to do this thing inside - if done in front of each request -> ping or try to run the query and, in the case of a reply containing a "gone away" is the second request and reconnect? If you do -> ping () - it should be very slow - better to make your wrapper, or is there something good and ready?

Separately - what is a good alternative to DBIx:: Connector, if the script can forkatsya, run under mod_perlom or something else? In DBIx:: Connector documentation such monstrous structures that ipolzovat it is not desirable.

Preamble

Preamble

perl -nle ' print $1 while m!(http://example.com/[^" >]+)!g'

<a href="http://example.com/one"> one </ a> and <a href="http://example.com/two"> two </ a>
http://example.com/first and http://example.com/second
none

http://example.com/one
http://example.com/two
http://example.com/first
http://example.com/second