Beefy Boxes and Bandwidth Generously Provided by pair Networks
Welcome to the Monastery
 
PerlMonks  

(crazyinsomniac) Re: Extract info from HTML

by crazyinsomniac (Prior)
on Nov 12, 2001 at 12:46 UTC ( #124774=note: print w/ replies, xml ) Need Help??


in reply to Extract info from HTML

Weeel I started writing code on this about two hours ago, and didn't do much until I actually started thinking about the problem some 15 minutes ago (when I put down the peanuts and exausted me votes for the day) and this is what I came up with:

First, I needed to pick a module to use, and HTML::TokeParser sat really well with me. The initial problem for me, was to figure out what "html" is the one I wan't, and I did what I always do when diagnosing such a problem, I dump the entire document token by token, in this case, with:

#!/usr/bin/perl -w use strict; use LWP::Simple; use HTML::TokeParser; my $url ="http://perlmonks.org/index.pl?node_id=110166"; my $rawHTML = get($url); # attempt to d/l the page to mem die "LWP::Simple messed up $!" unless ($rawHTML); my $tp; $tp = HTML::TokeParser->new(\$rawHTML) or die "WTF $tp gone bad: $!"; # And now -- a generic HTML::TokeParser loop while (my $token = $tp->get_token) { my $ttype = shift @{ $token }; print "TYPE : $ttype\n####\n"; printf( join( '', map { "$_:%s\n####\n" } 1..@{$token} ) , @{$token} ); print "####################################################\n\n"; } __END__ Which produces something like: TYPE : D #### 1:<!DOCTYPE HTML PUBLIC "-//W3C//DTD HTML 4.0 Transitional//EN" > #### #################################################### TYPE : T #### 1: #### 2: #### #################################################### TYPE : C #### 1:<!--took this out for IE6ites "http://www.w3.org/TR/REC-html40/loos +e.dtd"--> #### #################################################### TYPE : T #### 1: #### 2: #### #################################################### TYPE : S #### 1:html #### 2:HASH(0x1afeee0) #### 3:ARRAY(0x1afeef8) #### 4:<HTML> #### #################################################### TYPE : T #### 1: #### 2: #### ####################################################
Then, after "visualizing" what criteria I can use to pick out the stuff I need (noted in __END__), I crafted me while loop like so:
#!/usr/bin/perl -w use strict; use LWP::Simple; use HTML::TokeParser; my $url ="http://perlmonks.org/index.pl?node_id=110166"; my $rawHTML = get($url); # attempt to d/l the page to mem die "LWP::Simple messed up $!" unless ($rawHTML); my $tp; $tp = HTML::TokeParser->new(\$rawHTML) or die "WTF $tp gone bad: $!"; # And now -- a generic HTML::TokeParser loop while (my $token = $tp->get_token) { my $ttype = shift @{ $token }; if($ttype eq "S" and $token->[0] eq "br") { my ( @t ) = ( undef, #$tp->get_token, #S 0 $tp->get_token, #T 1 $tp->get_token, #S 2 $tp->get_token, #T 3 $tp->get_token, #E 4 $tp->get_token, #T 5 ); if( # ($t[0][0] eq "S" and $t[0][1] eq "br") and ($t[1][0] eq "T" and $t[1][1] =~ /by/) and ($t[2][0] eq "S" and $t[2][1] eq "a") and ($t[3][0] eq "T" ) and ($t[4][0] eq "E" and $t[4][1] eq "a") and ($t[5][0] eq "T" and $t[5][1] =~ /on \w{3} \d{2}, \d{4} at +/) ) { print $t[2][4], $t[3][1], $t[4][2], " | "; } } } # endof while (my $token = $p->get_token) undef $rawHTML; # no more raw html undef $tp; # destroy the HTML::TokeParser object (don't need it n +o more) __END__ ######### WITH ADDED NEWLINES FOR READABILITY AT >< <TR BGCOLOR=eeeeee><TD colspan=2> <UL> <font size=2> <A HREF="/index.pl?node_id=110247&lastnode_id=110166"> Re: Re: Name Space </A> <BR> by <A HREF="/index.pl?node_id=85506&lastnode_id=110166"> Hofmator </A> on Sep 05, 2001 at 02:27 </UL> </font></TD></tr> ########## BROKEN DOWN BY TOKEN TYPE : S #### 1:br #### 2:HASH(0x1af8128) #### 3:ARRAY(0x1afeeec) #### 4:<BR> #### #################################################### TYPE : T #### 1: by #### 2: #### #################################################### TYPE : S #### 1:a #### 2:HASH(0x1ab4384) #### 3:ARRAY(0x1ab6324) #### 4:<A HREF="/index.pl?node_id=85506&lastnode_id=110166"> #### #################################################### TYPE : T #### 1:Hofmator #### 2: #### #################################################### TYPE : E #### 1:a #### 2:</A> #### #################################################### TYPE : T #### 1: on Sep 05, 2001 at 02:27 #### 2: #### ####################################################
Which produced the following list:
japhy | Hofmator | tilly | davorg | scain | ichimunki | runrig | demerphq | merphq | shotgunefx | Masem | cLive ;-) | synapse0 | lo_tech | agent00013 | MrNobo1024 | Corion | demerphq | lo_tech | George_Sherston | Hofmator | Zaxo | idnopheq | dragonchild | herveus | wine | TheoPetersen | toadi | dga | mexnix | ybiC | {NULE} | theorbtwo | George_Sherston | Jouke | George_Sherston | tye | gregor42 | Guildenstern | sifukurt | CubicSpline | scain | zakzebrowski | jackdied | suaveant | poqui | mikeB | davis | s173451000 | blakem | George_Sherston | PotPieMan | mr_mischief | Zecho | earthboundmisfit | kwoff | Arguile | chaoticset | BrentDax | Aighearach | basicdez | brianarn | George_Sherston | BooK | riffraff | seanbo | Maestro_007 | stefan k | dthacker | Hero Zzyzzx | beretboy | Veachian64 | giulienk | blakem | George_Sherston |
The lesson here is, thank god vroom has a consistent format making it possible for me to decide what i want relatively easily (and thank god for HTML::TokeParser including the RAW html so I don't have to do much recreating, just repiecing ;D).

Is it elegant? I don't care, it makes sense to me (in practice and in theory).

update: oh yeah, it's not sorted, cause I don't actually "collect" the urls (users/userids) I want, cause like you can see, I just print them out.

This may help (a token can look like):

["S", $tag, $attr, $attrseq, $text] ["E", $tag, $text] ["T", $text, $is_data] ["C", $text] ["D", $text] ["PI", $token0, $text]
update: oh, point taken, that's just a simple oversight on my part, all i'd have to do is add a couple of more tokens... later ;D

 
___crazyinsomniac_______________________________________
Disclaimer: Don't blame. It came from inside the void

perl -e "$q=$_;map({chr unpack qq;H*;,$_}split(q;;,q*H*));print;$q/$q;"


Comment on (crazyinsomniac) Re: Extract info from HTML
Select or Download Code
Re: (crazyinsomniac) Re: Extract info from HTML
by George_Sherston (Vicar) on Nov 12, 2001 at 13:51 UTC
    Huh, I always thought Tokeparser was something to do with getting stoned in an orderly recursive fashion. :) That was interesting. Q: can you make it kick out nodes below a depth of 1? - otherwise it's an index of all nodes, rather than an index of nodes where people give their etymology... Q2 - can it pick up the node_id of the named monk's node in this thread, rather than the named monk's home node?

    George Sherston
      Well, even though this wasnt addressed to me:

      Mine will extract all the above information just change the following lines

      print "($depth)$monkname posted '$monkname' on $date\n"; $hashref->{$monkname}->{$node_id}={ date=>$date, title=>$title, depth=>$depth };
      Then you can extract whatever you want.
      $VAR1 = { 'demerphq' => { '110238' => { 'depth' => '13', 'title' => 'Corions Name Space +', 'date' => 'Sep 05, 2001 at 01: +04' }, 'Home' => '108447', '110195' => { 'depth' => '12', 'title' => 'Re: Name Space', 'date' => 'Sep 04, 2001 at 15: +46' } }, 'George_Sherston' => { 'Home' => '103111', '124767' => { 'depth' => '13', 'title' => 'Re: Re: Nam +e Space', 'date' => 'Nov 11, 2001 + at 22:33' }, 'Name Space' => { 'depth' => '9', 'title' => 'Name Sp +ace', 'date' => 'Sep 04, +2001 at 13:33' }, '121046' => { 'depth' => '14', 'title' => 'Re: Re: Re: + Name Space', 'date' => 'Oct 24, 2001 + at 01:21' }, '117665' => { 'depth' => '13', 'title' => 'Re: TheOrbT +wo\'s Name Space', 'date' => 'Oct 09, 2001 + at 00:05' }, '117303' => { 'depth' => '13', 'title' => 'Re: Re: Nam +e Space', 'date' => 'Oct 07, 2001 + at 03:57' }, '110244' => { 'depth' => '13', 'title' => 'Re: Re: Nam +e Space', 'date' => 'Sep 05, 2001 + at 01:58' }, '122854' => { 'depth' => '13', 'title' => 'Re: Re: Nam +e Space', 'date' => 'Nov 02, 2001 + at 08:07' } }, };
      Note that the depths are as follows:9 root node, 12 reply, 13, reply to a reply...
      But a thought: You dont want the posts from just a fixed depth in the parse tree. That would for instance eliminate you from the list (you dont have a reply to yourself) as well as anyone who explained their name in a reply to another persons explaination, merphq would be an example, however I believe there are more as well.

      Actually, one of the more interesting issues with this thread was acurately picking up all names from all levels, there is an annoying habit of <UL> tags messing up the pattern, also of the main post being marked up differently.

      Anyway, Ill revisit this a bit later, :-)

      Yves / DeMerphq
      --
      Have you registered your Name Space?

        Those whom the gods would destroy they first interest them in parsing natural language... really, in order to come up with a satisfactory solution to this, we're going to have to find a way to distinguish between the content of nodes... we need a script that can make an intelligent guess whether the node is a response or an etymology. This is a bit too rich for my blood, but I look forward to seeing it done :)

        George Sherston
        Well, to be mildly critical there are two (or more depending on how you look at it) scenarios where we can/need to extract information from. Yours only matches one fixed version (yes i know it was deliberate decision :-) now for the record (and in rehersal for that tutorial you suggested :-) Ill list the others:
        # Main node on page (Top most) <TD valign=middle> <H3>Name Space</H3> <FONT size=2> ' by ' <A HREF="/index.pl?node_id=103111&lastnode_id=110166">George_S +herston</A> ' on Sep 04, 2001 at 13:33' </FONT> </TD> # Primary Reply (crazyinsomniacs pattern) <TD colspan=2> <font size=2> <A HREF="/index.pl?node_id=110195&lastnode_id=110166">Re: Name + Space</A> <BR> ' by ' <A HREF="/index.pl?node_id=108447&lastnode_id=110166">demerphq +</A> ' on Sep 04, 2001 at 15:46' </font> </TD> # Note that the <UL> tag is incorrectly nested with regards to the <FO +NT> tag # Reply to a reply <TD colspan=2> <UL> <font size=2> <A HREF="/index.pl?node_id=110244&lastnode_id=110166">Re: +Re: Name Space</A> <BR> ' by ' <A HREF="/index.pl?node_id=103111&lastnode_id=110166">Geor +ge_Sherston</A> ' on Sep 05, 2001 at 01:58' </UL> </font> # Reply to a reply of a reply # each extra layer of depth has an extra <UL> tag inserted <TD colspan=2> <UL> <UL> <font size=2> <A HREF="/index.pl?node_id=121046&lastnode_id=110166"> +Re: Re: Re: Name Space</A> <BR> ' by ' <A HREF="/index.pl?node_id=103111&lastnode_id=110166"> +George_Sherston</A> ' on Oct 24, 2001 at 01:21' </UL> </UL> </font> </TD>
        Note the buggy HTML? :-)

        So what I did was look for the content of the FONT tag. If it matches a 'finger print' for one of the following two

        <font size=2> # Optional part begins <A HREF="/index.pl?node_id=121046&lastnode_id=110166">Re: Re: Re +: Name Space</A> <BR> # Optional part ends ' by ' <A HREF="/index.pl?node_id=103111&lastnode_id=110166">George_She +rston</A> ' on Oct 24, 2001 at 01:21' </font>
        Then I do a few more checks to make sure it isnt a spurious match, if they pass then I consider it the title/author/date of the node. A bit of extraction of the tags attributes and presto, we have the home node and post node ids. (With the exception of the main post, where we can only extract the title, not the ID)

        This would be sooooo much easier if there were class attributes in the tags, such as <TD class="post"> , but considering the buggy HTML, I suppose class attributes are low on the priority list. (BTW, cant wait to join the PM dev team, id like to have a crack at cleaning up some of the HTML, now that im getting into parsing it :-)

        Yves / DeMerphq
        --
        Have you registered your Name Space?

Log In?
Username:
Password:

What's my password?
Create A New User
Node Status?
node history
Node Type: note [id://124774]
help
Chatterbox?
and the web crawler heard nothing...

How do I use this? | Other CB clients
Other Users?
Others meditating upon the Monastery: (5)
As of 2014-12-22 05:14 GMT
Sections?
Information?
Find Nodes?
Leftovers?
    Voting Booth?

    Is guessing a good strategy for surviving in the IT business?





    Results (110 votes), past polls