Commit | Line | Data |
---|---|---|
6d72dad2 TH |
1 | #! /usr/bin/perl |
2 | # | |
3 | # parsedb.pl | |
4 | # | |
5 | # This script will parse a database with raw header information | |
6 | # from a INN feed to a structured database. | |
7 | # | |
8 | # It is part of the NewsStats package. | |
9 | # | |
10 | # Copyright (c) 2013 Thomas Hochstein <thh@inter.net> | |
11 | # | |
12 | # It can be redistributed and/or modified under the same terms under | |
13 | # which Perl itself is published. | |
14 | ||
15 | BEGIN { | |
16 | our $VERSION = "0.01"; | |
17 | use File::Basename; | |
18 | # we're in .../bin, so our module is in ../lib | |
19 | push(@INC, dirname($0).'/../lib'); | |
20 | } | |
21 | use strict; | |
22 | use warnings; | |
23 | ||
24 | use NewsStats qw(:DEFAULT :TimePeriods :SQLHelper); | |
25 | ||
26 | use DBI; | |
27 | use Getopt::Long qw(GetOptions); | |
28 | Getopt::Long::config ('bundling'); | |
29 | ||
9630376c TH |
30 | use Encode qw/decode/; |
31 | use Mail::Address; | |
32 | ||
6d72dad2 TH |
33 | ################################# Definitions ################################## |
34 | ||
35 | # define header names with separate database fields | |
36 | my %DBFields = ('date' => 'date', | |
37 | 'references' => 'refs', | |
38 | 'followup-to' => 'fupto', | |
39 | 'from' => 'from_', | |
40 | 'sender' => 'sender', | |
41 | 'reply-to' => 'replyto', | |
42 | 'subject' => 'subject', | |
43 | 'organization' => 'organization', | |
44 | 'lines' => 'linecount', | |
45 | 'approved' => 'approved', | |
46 | 'supersedes' => 'supersedes', | |
47 | 'expires' => 'expires', | |
48 | 'user-agent' => 'useragent', | |
49 | 'x-newsreader' => 'xnewsreader', | |
50 | 'x-mailer' => 'xmailer', | |
51 | 'x-no-archive' => 'xnoarchive', | |
52 | 'content-type' => 'contenttype', | |
53 | 'content-transfer-encoding' => 'contentencoding', | |
54 | 'cancel-lock' => 'cancellock', | |
55 | 'injection-info' => 'injectioninfo', | |
56 | 'x-trace' => 'xtrace', | |
57 | 'nntp-posting-host' => 'postinghost'); | |
58 | ||
59 | # define field list for database | |
60 | my @DBFields = qw/day mid refs date path newsgroups fupto from_ from_parsed | |
61 | from_name from_address sender sender_parsed sender_name | |
62 | sender_address replyto replyto_parsed replyto_name | |
63 | replyto_address subject subject_parsed organization linecount | |
64 | approved supersedes expires useragent xnewsreader xmailer | |
65 | xnoarchive contenttype contentencoding cancellock injectioninfo | |
66 | xtrace postinghost headers disregard/; | |
67 | ||
68 | ################################# Main program ################################# | |
69 | ||
70 | ### read commandline options | |
71 | my ($OptDay,$OptDebug,$OptParseDB,$OptRawDB,$OptTest,$OptConfFile); | |
72 | GetOptions ('d|day=s' => \$OptDay, | |
73 | 'debug!' => \$OptDebug, | |
74 | 'parsedb=s' => \$OptParseDB, | |
75 | 'rawdb=s' => \$OptRawDB, | |
76 | 't|test!' => \$OptTest, | |
77 | 'conffile=s' => \$OptConfFile, | |
78 | 'h|help' => \&ShowPOD, | |
79 | 'V|version' => \&ShowVersion) or exit 1; | |
80 | ||
81 | ### read configuration | |
82 | my %Conf = %{ReadConfig($OptConfFile)}; | |
83 | ||
84 | ### override configuration via commandline options | |
85 | my %ConfOverride; | |
86 | $ConfOverride{'DBTableRaw'} = $OptRawDB if $OptRawDB; | |
87 | $ConfOverride{'DBTableParse'} = $OptParseDB if $OptParseDB; | |
88 | &OverrideConfig(\%Conf,\%ConfOverride); | |
89 | ||
90 | ### get time period | |
91 | ### and set $Period for output and expression for SQL 'WHERE' clause | |
92 | my ($Period,$SQLWherePeriod) = &GetTimePeriod($OptDay,'day'); | |
93 | # bail out if --month is invalid or "all" | |
94 | &Bleat(2,"--day option has an invalid format - please use 'YYYY-MM-DD' or ". | |
95 | "'YYYY-MM-DD:YYYY-MM-DD'!") if (!$Period or $Period eq 'all time'); | |
96 | ||
97 | ### init database | |
98 | my $DBHandle = InitDB(\%Conf,1); | |
99 | ||
100 | ### get & write data | |
101 | &Bleat(1,'Test mode. Database is not updated.') if $OptTest; | |
102 | ||
103 | # create $SQLWhereClause | |
104 | my $SQLWhereClause = SQLBuildClause('where',$SQLWherePeriod,'NOT disregard'); | |
105 | ||
106 | # delete old data for current period | |
107 | if (!$OptTest) { | |
108 | print "----------- Deleting old data ... -----------\n" if $OptDebug; | |
109 | my $DBQuery = $DBHandle->do(sprintf("DELETE FROM %s.%s %s", | |
110 | $Conf{'DBDatabase'},$Conf{'DBTableParse'}, | |
111 | $SQLWhereClause)) | |
112 | or &Bleat(2,sprintf("Can't delete old parsed data for %s from %s.%s: ". | |
113 | "$DBI::errstr\n",$Period, | |
114 | $Conf{'DBDatabase'},$Conf{'DBTableParse'})); | |
115 | }; | |
116 | ||
117 | # read from DBTableRaw | |
118 | print "-------------- Reading data ... -------------\n" if $OptDebug; | |
119 | my $DBQuery = $DBHandle->prepare(sprintf("SELECT id, day, mid, peer, path, ". | |
120 | "newsgroups, headers, disregard ". | |
121 | "FROM %s.%s %s", $Conf{'DBDatabase'}, | |
122 | $Conf{'DBTableRaw'}, $SQLWhereClause)); | |
123 | $DBQuery->execute() | |
124 | or &Bleat(2,sprintf("Can't get data for %s from %s.%s: ". | |
125 | "$DBI::errstr\n",$Period, | |
126 | $Conf{'DBDatabase'},$Conf{'DBTableRaw'})); | |
127 | ||
128 | # set output and database connection to UTF-8 | |
129 | # as we're going to write decoded header contents containing UTF-8 chars | |
130 | binmode(STDOUT, ":utf8"); | |
131 | $DBHandle->do("SET NAMES 'utf8'"); | |
132 | ||
133 | # parse data in a loop and write it out | |
134 | print "-------------- Parsing data ... -------------\n" if $OptDebug; | |
135 | while (my $HeadersR = $DBQuery->fetchrow_hashref) { | |
136 | my %Headers = %{$HeadersR}; | |
137 | ||
138 | # parse $Headers{'headers'} ('headers' from DBTableRaw) | |
139 | # merge continuation lines | |
140 | # from Perl Cookbook, 1st German ed. 1999, pg. 91 | |
141 | $Headers{'headers'} =~ s/\n\s+/ /g; | |
142 | # split headers in single lines | |
143 | my $OtherHeaders; | |
144 | for (split(/\n/,$Headers{'headers'})) { | |
145 | # split header lines in header name and header content | |
146 | my ($key,$value) = split(/:/,$_,2); | |
147 | $key =~ s/\s*//; | |
148 | $value =~ s/^\s*(.+)\s*$/$1/; | |
149 | # save each header, separate database fields in %Headers, | |
150 | # the rest in $OtherHeaders (but not Message-ID, Path, Peer | |
151 | # and Newsgroups as those do already exist) | |
152 | if (defined($DBFields{lc($key)})) { | |
153 | $Headers{$DBFields{lc($key)}} = $value; | |
154 | } else { | |
155 | $OtherHeaders .= sprintf("%s: %s\n",$key,$value) | |
156 | if lc($key) !~ /^(message-id|path|peer|newsgroups)$/; | |
157 | } | |
158 | } | |
159 | # replace old (now parsed) $Headers{'headers'} with remanining $OtherHeaders | |
160 | chomp($OtherHeaders); | |
161 | $Headers{'headers'} = $OtherHeaders; | |
162 | ||
9630376c TH |
163 | foreach ('from_','sender', 'replyto', 'subject') { |
164 | if ($Headers{$_}) { | |
165 | my $HeaderName = $_; | |
166 | $HeaderName =~ s/_$//; | |
167 | # decode From: / Sender: / Reply-To: / Subject: | |
168 | if ($Headers{$_} =~ /\?(B|Q)\?/) { | |
169 | $Headers{$HeaderName.'_parsed'} = decode('MIME-Header',$Headers{$_}); | |
170 | } | |
171 | # extract name(s) and mail(s) from From: / Sender: / Reply-To: | |
172 | # in parsed form, if available | |
173 | if ($_ ne 'subject') { | |
174 | my @Address; | |
175 | # start parser on header or parsed header | |
176 | # @Address will have an array of Mail::Address objects, one for | |
177 | # each name/mail (you can have more than one person in From:!) | |
178 | if (defined($Headers{$HeaderName.'_parsed'})) { | |
179 | @Address = Mail::Address->parse($Headers{$HeaderName.'_parsed'}); | |
180 | } else { | |
181 | @Address = Mail::Address->parse($Headers{$_}); | |
182 | } | |
183 | # split each Mail::Address object | |
184 | foreach (@Address) { | |
185 | # take address part | |
186 | $Headers{$HeaderName.'_address'} = $_->address(); | |
187 | # take name part form "phrase", if there is one: | |
188 | # From: My Name <addr@ess> (Comment) | |
189 | # otherwise, take it from "comment": | |
190 | # From: addr@ess (Comment) | |
191 | $Headers{$HeaderName.'_name'} = $_->comment() | |
192 | unless $Headers{$HeaderName.'_name'}= $_->phrase; | |
193 | $Headers{$HeaderName.'_name'} =~ s/^\((.+)\)$/$1/; | |
194 | # FIMXE - handle more than one Mail::Address object! | |
195 | } | |
196 | } | |
197 | } | |
198 | } | |
199 | ||
6d72dad2 TH |
200 | # order output for database entry: fill @SQLBindVars |
201 | print "-------------- Next entry:\n" if $OptDebug; | |
202 | my @SQLBindVars; | |
203 | foreach (@DBFields) { | |
204 | if (defined($Headers{$_}) and $Headers{$_} ne '') { | |
205 | push (@SQLBindVars,$Headers{$_}); | |
206 | printf ("FOUND: %s -> %s\n",$_,$Headers{$_}) if $OptDebug; | |
207 | } else { | |
208 | push (@SQLBindVars,undef); | |
209 | } | |
210 | } | |
211 | ||
212 | # write data to DBTableParse | |
213 | if (!$OptTest) { | |
214 | print "-------------- Writing data ... -------------\n" if $OptDebug; | |
215 | my $DBWrite = | |
216 | $DBHandle->prepare(sprintf("INSERT INTO %s.%s (%s) VALUES (%s)", | |
217 | $Conf{'DBDatabase'}, | |
218 | $Conf{'DBTableParse'}, | |
219 | # get field names from @DBFields | |
220 | join(', ',@DBFields), | |
221 | # create a list of '?' for each DBField | |
222 | join(', ', | |
223 | split(/ /,'? ' x scalar(@DBFields))) | |
224 | )); | |
225 | $DBWrite->execute(@SQLBindVars) | |
226 | or &Bleat(2,sprintf("Can't write parsed data for %s to %s.%s: ". | |
227 | "$DBI::errstr\n",$Period, | |
228 | $Conf{'DBDatabase'},$Conf{'DBTableParse'})); | |
229 | $DBWrite->finish; | |
230 | } | |
231 | }; | |
232 | $DBQuery->finish; | |
233 | ||
234 | ### close handles | |
235 | $DBHandle->disconnect; | |
236 | ||
237 | print "------------------- DONE! -------------------\n" if $OptDebug; | |
238 | __END__ | |
239 | ||
240 | ################################ Documentation ################################# | |
241 | ||
242 | =head1 NAME | |
243 | ||
244 | parsedb - parse raw data and save it to a database | |
245 | ||
246 | =head1 SYNOPSIS | |
247 | ||
248 | B<parsedb> [B<-Vht>] [B<--day> I<YYYY-MM-DD> | I<YYYY-MM-DD:YYYY-MM-DD>] [B<--rawdb> I<database table>] [B<--parsedb> I<database table>] [B<--conffile> I<filename>] [B<--debug>] | |
249 | ||
250 | =head1 REQUIREMENTS | |
251 | ||
252 | See L<doc/README>. | |
253 | ||
254 | =head1 DESCRIPTION | |
255 | ||
256 | ... | |
257 | ||
258 | =head2 Configuration | |
259 | ||
260 | ... | |
261 | ||
262 | =head1 OPTIONS | |
263 | ||
264 | =over 3 | |
265 | ||
266 | =item B<-V>, B<--version> | |
267 | ||
268 | Print out version and copyright information and exit. | |
269 | ||
270 | =item B<-h>, B<--help> | |
271 | ||
272 | Print this man page and exit. | |
273 | ||
274 | =item B<--debug> | |
275 | ||
276 | Output (rather much) debugging information to STDOUT while processing. | |
277 | ||
278 | =item B<-t>, B<--test> | |
279 | ||
280 | Do not write results to database. You should use B<--debug> in | |
281 | conjunction with B<--test> ... everything else seems a bit pointless. | |
282 | ||
283 | =item B<-d>, B<--day> I<YYYY-MM-DD[:YYYY-MM-DD]> | |
284 | ||
285 | Set processing period to a single day in YYYY-MM-DD format or to a time | |
286 | period between two days in YYYY-MM-DD:YYYY-MM-DD format (two days, separated | |
287 | by a colon). | |
288 | ||
289 | Defaults to yesterday. | |
290 | ||
291 | =item B<--rawdb> I<table> (raw data table) | |
292 | ||
293 | Override I<DBTableRaw> from F<newsstats.conf>. | |
294 | ||
295 | =item B<--parsedb> I<table> (parsed data table) | |
296 | ||
297 | Override I<DBTableParse> from F<newsstats.conf>. | |
298 | ||
299 | =item B<--conffile> I<filename> | |
300 | ||
301 | Load configuration from I<filename> instead of F<newsstats.conf>. | |
302 | ||
303 | =back | |
304 | ||
305 | =head1 INSTALLATION | |
306 | ||
307 | See L<doc/INSTALL>. | |
308 | ||
309 | =head1 EXAMPLES | |
310 | ||
311 | ... | |
312 | ||
313 | =head1 FILES | |
314 | ||
315 | =over 4 | |
316 | ||
317 | =item F<bin/parsedb.pl> | |
318 | ||
319 | The script itself. | |
320 | ||
321 | =item F<lib/NewsStats.pm> | |
322 | ||
323 | Library functions for the NewsStats package. | |
324 | ||
325 | =item F<etc/newsstats.conf> | |
326 | ||
327 | Runtime configuration file. | |
328 | ||
329 | =back | |
330 | ||
331 | =head1 BUGS | |
332 | ||
333 | Please report any bugs or feature requests to the author or use the | |
334 | bug tracker at L<http://bugs.th-h.de/>! | |
335 | ||
336 | =head1 SEE ALSO | |
337 | ||
338 | =over 2 | |
339 | ||
340 | =item - | |
341 | ||
342 | L<doc/README> | |
343 | ||
344 | =item - | |
345 | ||
346 | L<doc/INSTALL> | |
347 | ||
348 | =back | |
349 | ||
350 | This script is part of the B<NewsStats> package. | |
351 | ||
352 | =head1 AUTHOR | |
353 | ||
354 | Thomas Hochstein <thh@inter.net> | |
355 | ||
356 | =head1 COPYRIGHT AND LICENSE | |
357 | ||
358 | Copyright (c) 2013 Thomas Hochstein <thh@inter.net> | |
359 | ||
360 | This program is free software; you may redistribute it and/or modify it | |
361 | under the same terms as Perl itself. | |
362 | ||
363 | =cut |