0.35 26 April 2002
Updated some omissions to the NexTrieve.pm documentation.
Added scripts "targz_collect" and "targz_count".
Fixed errors caused by differently operating "pdftotext" program on
some systems in the test-suite of PDF.pm.
Fixed problem with new default case of "add_file" of Targz.pm.
0.34 25 April 2002
Added default case to "add_file" of Targz.pm to more easily handle
incoming mail messages.
5 April 2002
Changed some documentation after discussion with Mark Overmeer at the
Amsterdam.pm meeting.
0.33 4 April 2002
Fixed some annoying errors when manifying Text sequence by
changing that to 1234 in HTML.pm, RFC822.pm and Message.pm.
Added mime-handler "_pdf" to MIME.pm for handling "application/pdf"
MIME-types of RFC822.pm, Message.pm and Mbox.pm indirectly. This means
that emails with PDF-files attach will now also index the PDF-files.
First releasable version of PDF.pm completed including (limited)
test-suite (t/18pdf.t). Also added "pdf2ntvml" script plus test-suite
(t/75pdf.t).
Changed "add_news" and "_resync_news" methods in Targz.pm to allow
for automatic recovery from a Net::NNTP object that has gone stale.
2 April 2002
Added "_fetch_file" method to NexTrieve.pm for fetching data as an
external file. Added "DESTROY" method to NexTrieve.pm for
automatically removing temporary files added by _fetch_file and
possibly others in the future.
Commenced work on PDF.pm, based on "pdfinfo" and "pdftotext" programs
of the xpdf package, located at http://www.foolabs.com/xpdf/ . Added
all the hooks and documentation in associated packages.
0.32 1 April 2002
Fixed problem in method "ResourceFromIndex" in Index.pm. Some versions
of NexTrieve give error message that would trigger the "ok" check.
This is now fixed.
Changed method "_create_tarfile" in Targz.pm to first create the
tarfile and then gzip it. This approach allows incremental updates
of the tarfile, allowing unlimited number of files to be added to the
tarfile (it would bomb on huge numbers of messages in a single day
before). Adapted documentation to indicate a "gzip" program with the
"--best" parameter is also needed. This should probably lead to
better compression of the gzipped tarfiles.
31 March 2002
Some more tuning in "_resync_news" of Targz.pm. Now correctly handles
the case with a lot of missing messages: if the date of a message is
two days or more before the last date of a message, then a collect is
started from the message after that message.
25 March 2002
Fixed a small problem in internal "_resync_news" method of Targz.pm
that would loop on missing messages in the target zone.
0.31 25 March 2002
Refined the internal "_resync_news" method to quickly handle "holes"
in the message stream. Now also uses a binary chop approach to find
the last message that's on the news server that is already in the
targz. This all applies to Targz.pm of course.
24 March 2002
Added and documented method "add_news" to Targz.pm. Takes a Net::NNTP
object and reads messages from there, adding them to the targz.
Handles re-syncing with newsgroups by a mix of date and message-id
checks.
Added and documented method "name" to Targz.pm. Added and documented
method "count_storable", which is the same as "count" but uses the
Storable module for persistency to prevent having to unpack tarfiles
that haven't changed. Added checks to test-suite.
Modified internal method "auto_clean" to "no_auto_clean" and documented
it. Modified internal method "clean" to only work as an object method
and documented it. Both in Targz.pm.
Simplified some internals in Targz.pm. The tar program must now also
be able to handle the "--directory" directive.
23 March 2002
Added and documented a "tarfile" method to Targz.pm.
Made the datestamp checking routine in Targz.pm a little smarter so
that it now also recognizes and handles NNTP-Posting-Date: and
X-Trace: headers.
Added support for an external hash to "count" method of Targz.pm:
using an external hash can make things a lot faster because it does
not need to read tar-files that haven't changed.
Made directory parameter to Targz method of NexTrieve.pm default to
the current directory.
22 March 2002
Adapted the undocumented "files" method of Docseq.pm so that it can
accept a processor routine parameter. Also documented the method now.
It is now useful as a basic conversion feature for any type of
conversion by other modules.
0.30 18 March 2002
First version of Targz.pm completed including documentation. You
can now quickly store both messages as well as unix mailboxes in the
NexTrieve::Targz archive format.
Added return value for success to method "splat" in NexTrieve.pm.
Added "filename:id" feature to _fetch_content_from_filename in
NexTrieve.pm, allowing filenames to be specified with an ":id"
suffix, which would then fill the "id" key in the content hash.
So you can now specify an absolute (temporary) filename with an
ID specification in one go. This applies to RFC822.pm and HTML.pm
Feature created to fix the re-XMLing process of Targz.pm.
17 March 2002
First version of Targz.pm almost ready. Only a few cleanup issues to
be fixed.
Create specific method "write_file" to Document.pm so that the encoding
information is saved when a single document is written out. All
other methods to get at the XML of a Document object still return
the XML _without_ the processor instruction for easy inclusion in
document sequences.
Added dependency on Cwd and File::Copy to Makefile.PL. Needed for
Targz.pm.
Added additional key-value pairs specification to the Document method
of the RFC822.pm. Needed for Targz.pm.
16 March 2002
Started work on Targz.pm based on the scripts developed the past year.
Bolted dependency for IO::File, IO::Socket and Date::Parse into
NexTrieve.pm. They seem to have been around forever: no need for
cleverness there.
0.29 11 March 2002
Some documentation fixes to Message.pm and NexTrieve.pm. Renamed
Overview.pod back to Overview.pm, as that _will_ show up for reading
on the various CPAN related websites.
0.28 11 March 2002
Finished initial version of Message.pm after some more discussions with
Mark Overmeer. There doesn't seem to be a need for a Mail::Box
interface yet, so that source will be dumped now.
Changed Overview.pm to Overview.pod.
10 March 2002
Created MIME.pm module as a stash for MIME-conversion routines.
Adapted RFC822.pm so that it uses the new MIME.pm module, removed its
own versions of _plain and _html.
Started work on Message.pm for converting Perl Mail::Message objects to
document sequences. Added test-suite for it as well. Initially
developed as NexTrieve::Mail::Box.pm, but this turned out to be too
much double work. After discussions with Mark Overmeer, the author
of Mail::Box and Mail::Message, it seemed to make much more sense to
interface at the message level rather than at the mailbox level.
Oops. Lost the NAME and SYNOPSIS section in Overview.pm while
copying the text that was made off-line. Restored again now. This
caused the Overview.pm to become "invisible" on CPAN, which is a
pity for a module that consists of documentation only.
0.27 9 March 2002
Added documentation for methods "texttype" and "texttypes" to the
Query.pm module: they were missing.
Added Overview.pm documentation module. Moved some of the
documentation from NexTrieve.pm to it.
0.26 6 March 2002
Finished first complete documentation of Resource.pm.
Removed the "basedir" method from Resource.pm. The NexTrieve "basedir"
feature is on the way out and shouldn't have existed in the Perl
modules in the first place. Needed to adapt quite some tests in the
test-suite as they used "basedir" as an example method.
0.25 5 March 2002
Finished first complete documentation of Query.pm, Querylog.pm,
Replay.pm and Search.pm.
Added Query method to Replay.pm.
Added documentation for "ampersandize" and "normalize" to NexTrieve.pm.
0.24 4 March 2002
Finished first complete documentation of Docseq.pm, Document.pm,
Hitlist.pm, Hitlist::Hit.pm, Index.pm, Mbox.pm.
Changed method "ResourceFromIndex" in Index.pm to use "ntvcheck" rather
than "ntvopt": the --xml functionality should be there. Adapted
test-suite so it now correctly handles the absence of --xml
functionality in ntvcheck.
3 March 2002
Finished first complete documentation of Daemon.pm.
Adapted method "executable" in NexTrieve.pm to return the license
expiration info as a datestamp: YYYYMMDD.
Changed method "PrintError" in NexTrieve.pm to accept the "cluck"
keyword. If specified, the $SIG{__WARN__} handler is set to
Carp::cluck.
Changed method "RaiseError" in NexTrieve.pm to accept the "confess"
keyword. If specified, the $SIG{__DIE__} handler is set to
Carp::confess.
Changed method "ResourceFromIndex" in Index.pm to use "ntvopt" rather
than "ntvcheck": the --xml functionality seems to have moved.
Finished first complete documentation of DBI.pm.
Finished first complete documentation of RFC822.pm.
Removed "use NexTrieve::Resource" from HTML.pm and RFC822.pm. They
are only needed when the "Resource" method would be called, which is
not too often. The NexTrieve::Resource module must now be explicitely
specified in the "use NexTrieve qw()" list when needed. Adapted the
test-suite accordingly.
Added "mailsimple" method to RFC822.pm. Same as default settings of
the "mailbox2ntvml" script.
Finished first complete documentation of HTML.pm.
Added "embed" to _default_removecontainers in NexTrieve.pm.
Minor fix to _intext_recode of NexTrieve.pm to handle the case when
no input is given. This was causing a lot of warnings in the
test-suite if MIME::xxx were not installed.
Minor fix to _plain and _html in RFC822.pm to allow handling of empty
text and html (which could be caused by MIME::Base64 and
MIME::QuotedPrint not being installed).
Added support for handling the case when MIME::Base64 and
MIME::QuotedPrint are not installed. They were handled by the modules
already, but not in the test-suite, causing errors when they shouldn't.
28 February 2002
First half of more complete documentation of HTML.pm.
0.23 28 February 2002
Added flag to internal method "_recoding_error" so that a different
error message is displayed when some data was actually returned.
Adapted method "_iconv" to use this new feature.
Changed handling of calling external "iconv" from a piped open to a
system with temporary input and output files. Apparently, that is the
only way to reliably obtain exit codes from iconv in older versions
of Perl.
Changed the handling of recoding =?encoding?Q?string?= strings inside
strings to _process_container. This should make the handling much
more general, and possibly less CPU-intensive as it is only done on
elements from the content-hash that are actually converted to
attributes or texttypes. Added "t/headerenc.mbox" and "t/asia.mbox"
test-cases.
0.22 27 February 2002
Added "archive" method to Mbox.pm. When an archive is specified, it
is assumed to be either a handle or a filename to be opened for
appending. Just before a message is processed, it will be written
to the archive, allowing developers to use this for a simple mail
archiving system. Added t/74mbox.t test for this functionality.
Fixed bug in Mbox that would occur if the same $docseq would be used
in multiple runs togethev with a conceptualmailbox and a baseoffset.
The second run, the baseoffset of the first run would be used. Now
the baseoffset is updated in the object after a run when a conceptual
mailbox is used.
Changed Mbox.pm also so that a conceptualmailbox is just that and that
you need to specify an offset in that case (if it's different from 0
that is). Adapted t/14mbox.t accordingly.
Made the use of -o obligatory when using -c. No longer looks up
offset assuming conceptualmailbox is a real file somewhere. Adapted
test-suite t/72mbox.t accordingly. This was in "mailbox2ntvml" of
course.
Fixed minor nit in "mailbox2ntvml": if defined($baseoffset) was not
needed at all.
0.21 26 February 2002
Fixed problem in the "mailbox2ntvml" script that would ignore the
-o (baseoffset) parameter. Added two test-suites for checking the
functionality of the -c and -o parameters of that script.
Added script "dbi2ntvml" for executing a query in a database and
having a document sequence created for the result.
Fixed problems with broken attachments that don't finish with a newline
in RFC822.pm by fixing the "next" and "nextnonewline" of the hidden
NexTrieve::handle object in NexTrieve.pm. Added a test-file
"badmime.mbox" to test for this eventuality.
Fixed problem in scripts "mailbox2ntvml" and "html2ntvml": the -E flag
for specifying the default input encoding, did not work. The default
input encoding was always set to 'iso-8859-1'.
Further refined the ucs-4 and ucs-2 encoding issues: made the
"utf3216check" method a lot smarter. It is now able to detect big
and little endian and sets the encoding information appropriately.
Added support for "ucs-2le" and "ucs-4le" to UTF8.pm. Added heuristics
to _normalize_encoding to convert "utf-32" and "utf-16" to the
appropriate "ucs*" version. Added HTML-files with little-endian
2 and 4 byte encodings to the test-suite.
Removed "header2attribute" and "header2texttype" methods from
RFC822.pm. Instead, the inheritable "field2attribute" and
"field2texttype" should now be used. Changed the documentation, the
test-suite and scripts accordingly.
Changed name of "ShowErrorsAsWarnings" method in NexTrieve.pm to
"PrintError" to conform with the generally accepted way that the
"Perl" DBI.pm works. Changed all occurrences in the modules, scripts
and test-suite to reflect this change.
Changed name of "DieOnError" method in NexTrieve.pm to "RaiseError"
to conform with the generally accepted way that the "Perl" DBI.pm
works. Changed all occurrences in the modules, scripts and test-suite
to reflect this change.
Added NexTrieve::DBI.pm module for creating document sequences out of
DBI statement handles (actually, any object that has a method that can
be called repeatedly and which returns a reference to a hash). It
is now easy to create document sequences out of databases! Added
small test-suite for it: t/15dbi.t.
Moved "field2attribute" and "field2texttype" methods from HTML.pm to
NexTrieve.pm, so they can be inherited by DBI.pm and other modules.
Removed the methods from HTML.pm as they are now inherited.
Removed now obsolete "titlemax" method from RFC822.pm.
Found that documents encoded in utf-32 or utf-16 were not being handled
correctly by html2ntvml. Fixed this by adding a method "utf3216check"
to NexTrieve.pm that will check its input for utf-32 or utf-16
encoding (by checking the first 8, respectively 4 bytes of the text)
and convert that to utf-8 when deemed to be utf-32/utf-16. Added
call to this method to HTML.pm and added two test-cases, right out of
the standard Apache distribution, for these encodings. Added the
conversion from utf-32 and utf-16 (actually: ucs-2be and ucs4-be)
to UTF8.pm, so that these conversions are done internally.
0.20 25 February 2002
Generalize the handling of pairs in HTML.pm. Added
"author" and "generator" to the content hash as extra keys if
available. Other keys should now be trivial to add and should
possibly be customizable externally.
Sometimes the _iconv method of NexTrieve.pm seems to not be able to
create the file. It now silently exists without invoking _iconv.
Should probably be handled differently.
Added "x-mac-roman" and "windows-874" as a standard encoding that can
be handled by UTF8.pm. This should allow processing of most MAC
and some documents with Thai characters.
Added feature to _fetch_content in NexTrieve.pm that checks for
protocol-type specifications in the id specified and, if found,
forces a "URL" type fetch. This change allows URL's to be specified
on input anywhere, but most specifically in the "html2ntvml" script.
Fixed problem in _fetch_from_url in NexTrieve.pm that would cause
URL's of the form "http://www.nextrieve.com" (note the missing
slash at the end) to fail.
Removed some superfluous tables from NexTrieve.pm that weren't
necessary anymore.
Fixed baseoffset problem in script "mailbox2ntvml" if the referenced
mailbox file didn't exist. Also killed warning in that case in
HTML.pm.
Found one case of badly formatted HTML that exposed various
problems in the Document method of HTML.pm. Fixed the problems and
added a test-case for it in the test-suite. Fixed the same problems
in the HTML-attachment handling of RFC822.pm.
Changed method "tempfilename" in NexTrieve.pm to use the complete
hex address in the filename rather than just the numeric part.
Added iso-885\d-* as misspellings for iso-8859-* to _normalize_encoding
in NexTrieve.pm. Also added "html" as a misspelling for "iso-8859-1".
Added checks in the test-suite to test for these misspellings.
Added source specification to several error messages in HTML.pm.
Changed the "create_module" script so that the UTF-8 values are
generated at module creation time rather than when substituting the
values in strings. Updated UTF8.pm accordingly. Should make things
significantly faster.
0.19 24 February 2002
Added -a and -p flag to "html2ntvml" script to activate the ASP-style
and PHP-style tag removal.
Most of the test-suite scripts will now show the XML if there was an
unexpected XML found in any conversion.
Made the general conversion of containers somewhat stricter in HTML.pm
so that there is less chance of throwing away valuable stuff.
Added methods "asp" and "php" to add a pre-processor subroutine to
the HTML-object for removing ASP-style tags in the form <%...%> and
PHP-style tags in the form from the HTML. Added checks to
make sure that it works.
Generalized checking of t/70html.t and t/71mbox.t so that regular
expressions can be placed in the stderr file, allowing for natural
language independent checking of error messages. This change was
inspired by Arnaud ASSAD's report of a problem with a French "speaking"
iconv.
Completed first phase of more or less complete documentation of the
NexTrieve.pm module, including small descriptions of the input and
output parameters of methods, rather than just an example call.
Fixed problem with the "encoding" method of NexTrieve.pm: setting an
encoding on an object that already has an encoding, now properly saves
the XML in the object of which the encoding was changed.
Added file VERSION so that stuff is easier to keep in CVS.
Added check for right version of modules to all of the scripts. Now,
a warning will be output if the script notices it is using a version
of the modules for which it was not designed.
Removed -c flag from call to "iconv": there are too many iconv's out
there that don't support it.
0.18 23 February 2002
Added "-c" flag to call to "iconv" so that it will not bomb on invalid
characters. Hopefully -c is valid to all versions of iconv out there.
Swiped iso-8859-* and windows-152* to UTF-8 conversion lists from the
Internet and created a conversion program that creates the source code
to the new NexTrieve::UTF8.pm module. From now on, all conversions
from iso-8859-* and windows-125* to UTF-8 are done natively, i.e.
without any external programs. Removed all the stuff related to
recoding that wasn't necessary anymore from NexTrieve.pm.
0.17 22 February 2002
Completely rewritten recoding in NexTrieve.pm. Lost the recoding hash
as well as the methods "_text_icon", "_default_recoding_handler",
"recode_handler" and "find_recoding". Instead of being recoding method
centric, a "from->to" centric approach has been taken. For each pair
of "from->to" recoding, a handler written in Perl is by default
available (e.g. for "iso-8859-1" to "utf-8"). If an encoding pair is
not found, first it is checked whether Text::Iconv can handle that
recoding. If so, a closure to the object doing that conversion is
created and saved. If that fails, a closure to an external "iconv"
program is created, using the generic "_iconv" method. This should
make recoding faster in many cases, and also handle dependencies on
external ways of doing recoding, much better.
Added some smart alecky way for RFC822.pm to allow the first attachment
to set the encoding of the document, rather than assuming iso-8859-1
and causing recodings to be done for windows-1252 attachments.
21 February 2002
Added stuff to NexTrieve.pm, HTML.pm, RFC822.pm and Mbox.pm so that
if there is a conversion error, the filename and line number (in case
of a mailbox) is shown in the error line.
Added conversion from "windows-1252" to "iso-8859-1" encoding to the
default recode handler in NexTrieve.pm.
Fixed problem with "Text::Iconv" recode handler if specified
directly rather than "found", in NexTrieve.pm.
Added some more checks to _normalize_encoding in NexTrieve.pm so that
"iso8859-1" and "iso_8859_1" are converted to "iso-8859-1". Added
some checks for this to t/01basic.t.
Added ^K as an extra null byte to be removed, in HTML.pm
20 February 2002
Removed character range 0x80-0x9f from illegal character range, as
these are valid windows-1252 characters and are no problem in
in iso-8859-1 even if they are supposed to be undefined.
Added _default_recoding_handler to NexTrieve.pm. This should be able
to convert from iso-8859-1 and windows-1252 to utf-8 by itself.
Allow this recoding method to be selected by the key "default".
Added a test file "win1252.html" to the test-suite.
Added ^L as an extra null byte to be removed, in HTML.pm
Fixed "find_recoding" to use the keys in the known recoding methods
hash.
0.16 20 February 2002
Adapted the check for an external "iconv" in NexTrieve.pm to do an
actual conversion, rather than checking for the -V flag. Should
really fix problem spotted by Nyk Cowham on a Mac OSX.
19 February 2002
Fixed problem in "xmllint" of NexTrieve.pm: value was being set even
if xmllint would not be available on a platform, causing the
test-suite to break. Spotted by Arnaud ASSAD.
Added method "shorten" to NexTrieve.pm for shortening strings and
making sure there are no broken entities at the end. Thought it would
be nice for processing routines, such as in "html2ntvml" script.
Since strings passed to processor routines are not normalized yet,
this is not a problem and for that reason this method is not needed.
Left in the source anyway as it seems to be a handy routine to have
anyway.
Fixed additional problem with HTML tag by changing the
behaviour of _process_container: now the normalization routine is _not_
passed as a parameter to the processing routine, but instead the
result of the processing routine is normalized before being put into
the XML stream. Added test-script t/70html.t for testing HTML files
with the "html2ntvml" script.
HTLM.pm now also removes ^Z as a null byte from the HTML stream before
processing: it appears that many Mac's and/or DOS editors add ^Z
characters at the end of the document: not removing them would cause
such documents be skipped if binary check is active.
0.15 18 February 2002
Fixed problem with containers appearing inside a <title> HTML tag in
HTML.pm. Title, keywords and description are now checked for
containers and removed as appropriate. Added a check to the test-suite
for this.
In NexTrieve/RFC822.pm the created document is immediately assumed to
be encoded in the DefaultInputEncoding unless there is a valid encoding
in the header. It no longer assumes the encoding of the first
processed attachment. This fixes a bug in the case when the recoding
of an attachment can not be done: before this would cause the whole
document to be skipped, now only the attachment in question will be
skipped.
The DefaultInputEncoding (in NexTrieve.pm) now defaults to "iso-8859-1"
even if never actually set. This causes a processor instruction to
_always_ become part of the XML when serialized and therefore needed
some changes to the test-suite.
In NexTrieve.pm, _normalize_encoding now changes any "us-ascii"
encoding name to "iso-8859-1", as "us-ascii" encoded texts in a
majority of cases include iso-8859-1 characters which would be
considered invalid with "us-ascii".
Wrapped opening of "iconv -V" in an eval to stop it from bombing if
no iconv is available, in NexTrieve.pm. Fixed after bug-report from
Nyk Cowham on a Mac OSX.
0.14 16 February 2002
Added new method "DefaultInputEncoding" in NexTrieve.pm. The value
of this method is now directly inherited by all the other modules.
Changed all the other modules to use $self->DefaultInputEncoding
rather than $self->NexTrieve->encoding.
Changed the way RFC822.pm reads a message to a nice hidden object
method of type NexTrieve::handle (as stored in NexTrieve.pm). This
should possibly fix the memory-hungryness for messages with large
attachments.
Changed the functionality of the "encoding" method: now if there is
an encoding already known for the object and a different encoding is
specified, then the XML will be serialised (if not already available)
and that XML will then be converted to the desired encoding. Added
a special version of the "encoding" method to Docseq.pm, as a Docseq
object can only be in UTF-8.
Changed all modules such that a Docseq object _always_ outputs the
serialised XML in UTF-8. Removed the -e parameter from the scripts
as these will always output in UTF-8 also.
In all situations where either content from a variable or a filename
could be specified, it is now possible to add one of more extra
parameters to indicate the type of content fetch. For the moment,
three types of content fetching are supported: '' for direct (value
is either the string or a reference to a list with a string, id and
epoch value), 'filename' to indicate the name of a file and 'url' to
indicate the content should be fetched from a URL. This is all based
on the content fetching mechanism in NexTrieve.pm.
Added documented but missing extra method setting functionality to
_new in Querylog.pm. Fixed problem in test for Querylog.pm in
t/82ntvsearchd.pm.
Added support for content fetching routines to NexTrieve.pm. Initial
base fetching routines are "_fetch_direct", "_fetch_from_filename" and
"_fetch_from_url". Added a central fetching method "_fetch_content".
Adapted "_filename_xml" to use this method of obtaining content, which
thus effectively allows this functionality from all module object
creation routines, such as $ntv->Resource.
0.13 16 February 2002
Moved character encoding issues from _process_part in RFC822.pm to the
mime-processor routines "_plain" and "_html". Adapted "_html" so that
it can work with HTML that specifies a different encoding as a <meta>
tag in the HTML from the one specified in the header. Added example
"bont.mbox" to list of tests.
Added support for binarycheck to RFC822.pm. Added support for -i flag
to mailbox2ntvml. Added example "ls.mbox" to list of tests.
Moved method "binarycheck" method from HTML.pm to NexTrieve.pm so that
it can be inherited by RFC822.pm.
Made sure no XML is returned from Document.pm if there is nothing
in it (before an empty <document> container would be returned). Fixed
test to reflect this new behaviour.
15 February 2002
Fixed warning in Docseq.pm if there was nothing to be piped.
12 February 2002
Added general method "xmllint" to NexTrieve.pm. When invoked with a
true value, will attempt to locate the program "xmllint" of the
libxml2 package. If found, any future actions that invoke
"write_string" either directly or indirectly (through an invocation
of "write_fh", "write_file" or "xml") will cause the generated XML
to be checked with the xmllint program and _if_ errors were found,
nullify the XML and add an error (with the error info from xmllint)
to the object. Mainly intended for internal debugging, but maybe
useful in other situations as well.
0.12 12 February 2002
Added -E flag to scripts docseq, mailbox2ntvml and html2ntvml to
allow specification of the default input encoding to be assumed in case
there is no other input encoding information available. Defaults to
"iso-8859-1".
Fixed conceptualmailbox functionality in script mailbox2ntvml and fixed
some warnings by properly initializing some variables in all scripts.
Added support for handling intext coded text in the form
=?iso-8859-2?Q?string=A9?=. to the headers in RFC822.pm and added a
test mbox for that case. Made small change to "recode" in NexTrieve.pm
to be able to support this.
Added method "bare" (for "bare XML") to Docseq.pm allowing the
<ntv:docseq> container to _not_ be emitted. Moved -b flag (binary
check) of html2ntvml script to -i. Added -b flag to docseq, html2ntvml
and mailbox2ntvml scripts.
Added general method "nopi" (for "no processor instruction") to
NexTrieve.pm. When applied to an object, it will cause the <?xml..>
to _not_ be emitted when XML is created for that object. Adapted the
docseq, html2ntvml and mailbox2ntvml scripts to allow for a -n flag
to omit the <?xml..?> processor instruction.
Fixed problem with dates not being processed in script/mailbox2ntvml
that was introduced yesterday as a result of some testing and the
Date::Parse absence fix.
0.11 11 February 2002
Fixed problem in "_iconv" of NexTrieve.pm. For some strange reason,
Perl would die if an encoding was encountered that was not supported
by iconv, even though the call was wrapped in an eval{}.
Checked all modules for calls to "openfile" and made sure that "slurp"
and "splat" were being used when appropriate. Also made sure that
when a file is being opened for reading, an explicit filemode is
specified.
Added method "splat" to NexTrieve.pm to write data to a handle and
then close the handle (the opposite of "slurp").
Added method "slurp" to NexTrieve.pm to read the entire contents of
an open handle. Adapted all modules that had the memory-hungry
structure with join( '',<$handle> ) to now use $self->slurp( $handle ).
Added check so that in all of the scripts, when they are fed with
something that doesn't look like a filename, it will produce a warning
rather than trying to open the string and possibly getting all sorts
of garbage on your file-system.
Fixed double escaping problem in NexTrieve.pm introduced earlier today.
Fixed test-suite problems in t/12html.t, t/13rfc822.t, t/14mbox.t and
t/71mbox.t that would occur if the Date::Parse module is not installed.
Fixed one more infinite loop problem in RFC822.pm when attempting to
decode faulty formed attachments.
Added new test-suite script t/71mbox.t for checking whether mails that
are known to produce problems in older versions, continue to be handled
correctly. Now 4 problem mails are in there: each test consists of
a sample mailbox (extension .mbox in the t directory) with a dummy
message preceding and following the actual message with a problem, as
well as a file with the expected stdout output (extension .stdout) as
well as a file with the expected stderr output (extension .stderr).
Adapted the MANIFEST accordingly. Currently 3 tests are being done
for each file: exit status, match on stdout output and match on stderr
output.
0.10 11 February 2002
Adapted HTML.pm to use the "_hashprocextra" method of NexTrieve.pm.
This simplified the "Document" method significantly.
Fixed warning message in NexTrieve::_iconv: if iconv failed to do
a conversion, don't bother trying to open the output file.
Implemented the content hash concept of HTML.pm into RFC822.pm as
well. This allows the "id" attribute to get another name and to
be missing from the XML at all if necessary. It also allows processing
routines to be assigned to the "id" attribute as well as for the
text (the '' empty attribute). Fixes problem in method "Resource"
which did not include the "id" attribute and was therefore out of
sync with the XML that was generated. Adapted the test-suite: some
order of the containers was changed as well as some whitespace
differences. Now also honours the "skip" method for skipping a
Document when so indicated inside a processing routine.
Moved (yet again) a lot of the intelligence of HTML.pm to NexTrieve.pm
in the "_hashprocextra" method, so that it can be used by both HTML.pm
and RFC822.pm and any other modules in the future (e.g. PDF.pm).
Adapted _add_container and _process_container to handle list references
(as used by RFC822.pm).
Changed all scripts in the "script" directory to use
"ShowErrorsAsWarnings" rather than "DieOnError". This should cause
the filters to continue even when there is a (simple) error such as
an attachment decoding error. Probably need something that allows
for finer tuning in the future.
Fixed problem in _process_parts of RFC822.pm that would cause an
infinite loop on faulty recursive attachments.
Changed "ResourceFromIndex" in Index.pm to handle garbage output in
older ntvopt's and no output in future ntvopt's.
10 February 2002
Wrapped "_iconv" conversion in an eval to prevent it from bombing Perl.
Added support for empty-tag processing routine for the rest HTML to be
processed and skip flag support to HTML.pm. This should now allow a
processing routine to process the HTML before creating the final XML
and to have any processing routine mark the document to be skipped
(e.g. after an MD5 check on the HTML reveals that there is already
a page with the same contents).
Added method "skip" to NexTrieve.pm as a generic way for processor
routines to indicate that the result of the processing should be
skipped.
Added support for no-name containers to _process_container and
_add_container in NexTrieve.pm.
9 February 2002
Added mask parameter to mkdir in t/80ntvindex.t and Index.pm: apparently
older versions of Perl 5 do not allow single argument mkdir().
Added some heuristics to _normalize_encoding of NexTrieve.pm to allow for broken encoding names such as "latin-1". Added test for this in
t/08docseq.t.
0.09 8 February 2002
Added methods "update_start" and "update_end" to Index.pm: this now
handles the creation of new versions of an index by first creating
a "indexdir.new" directory, adapting the Index object to have it index
in that directory, then when done indexing, move the current indexdir
to indexdir.old and moving indexdir.new to indexdir. Also copies
files in case of an incremental update. Still allows whatever way
you want for indexing. Removed the "Issue" idea from the TODO.
Added method "mkdir" to Index.pm to create the indexdir directory.
Changed class method "executable" in NexTrieve.pm to return the
program name as the first parameter instead of a flag, which is much
more handy. Adapted internal _command_log method to this
functionality as well as the ResourceFromIndex method in Index.pm.
Added method "restart" to Daemon.pm. Method "stop" now removes the
pid information from the object. Added test for this to
t/83ntvsearchd.t.
Made "stream" method of Docseq.pm default to STDOUT. Changed all the
scripts in the script directory to use that new feature.
Added check for extra attributes and texttypes to t/12html.t.
Final fix to ampersand: limit character number check to 3 digits
maximum to prevent overflow if number > 64K.
0.08 7 February 2002
Another fix to ampersand: now properly converts to instead of
&160;.
Made some of the XML creation less Perl version dependent by sorting
the keys in hashes where appropriate. Did the same with HTML.pm.
Fixes make test problems on older Perl versions but we probably should
find another way around this.
Fixed problem with -t parameter in "html2ntvml" script: was still
referencing the now non-existent "titlemax" method. Added an
attribute processor routine to fix the problem.
Fixed some documentation omissions in README and NexTrieve.pm pod.
6 February 2002
Fixed small problem in ampersand that would cause faulty entities such
as "word other word" to not convert to "word&160;other word".
Added "optimize" method to Index.pm. Added extra test-suite script
t/83ntvopt.t for checking ntvopt. NexTrieve::Index->executable now
allows filename parameter to check specific executablity of 'ntvopt'
or 'ntvidx-useopt.sh'.
Removed 2>/dev/null from the integrity check in NexTrieve.pm: we want
to know if something goes wrong.
0.07 6 February 2002
Added ResourceFromIndex method to Index.pm to create a Resource
object from an existing indexdir.
Added <A> as a default display container to NexTrieve.pm.
Added preprocessor concept to HTML.pm. Added "mhonarc" method that
sets up attributes, texttypes and processors for handling HTML-files
as generated by MHonArc. Added test-suite for MHonArc functionality.
Adapted test-suite for newer NexTrieve installations so that no -v
output from ntvindex is handled correctly.
Finished initial reconstruction of HTML.pm. Moved some more stuff from
RFC822.pm to NexTrieve.pm so that it can be used by HTML.pm as well.
Added "htmlsimple" method to HTML.pm so that you get the same behaviour
as before. Adapted script "html2ntvml" so that it used this
"htmlsimple" method to create same functionality.
5 February 2002
Continued work on HTML.pm. Removed "titlemax" method, as that should
now be handled by an attribute processing routine. Removed "key"
parameter from the API of processing routines: it did not make much
sense for RFC822 processing, it made even less sense for HTML
processing.
4 February 2002
Started work on HTML.pm to allow for extra attributes and texttypes,
and to have processor routines on attributes and texttypes. Changed
name of <filename> container to <id>, as that is more general. Method
"Document" also allows reference to list with ID and html to be passed
if both are in memory already.
Made checks on external modules Digest::MD5, Date::Parse and IO::Socket
the same: if they are already loaded when NexTrieve.pm is loaded, then
they will be activated immediately. Otherwise, they will be activated
on demand. This should give maximum flexibility (e.g. for a pre-
loading mod_perl environment) and minimum bloat (in on-demand
environments such as scripts).
Moved significant part of RFC822.pm intelligence to NexTrieve.pm, so
that it can also be inherited by HTML.pm and other modules in the
future.
3 February 2002
Changed RFC822.pm so that empty containers are not returned at all.
0.06 2 February 2002
Messed up an upload to CPAN, now it won't let me upload 0.05 again
properly, so bumped up the version to 0.06.
0.05 2 February 2002
Removed some debug crud from several tests.
Support for HTML in RFC822.pm now completed: if the message contains
HTML and not associated text, then the HTML will be stripped of its
containers and added as text. Added two more message with HTML checks
to the test-suite.
Removed 2>/dev/null from Index.pm and Daemon.pm so that any error
messages from NexTrieve will not be lost. Changed test-suite so that
when NexTrieve is installed, but a license can not be found, the tests
exit gracefully allowing an automatic install from CPAN in that case.
Create an "executable" class method in NexTrieve.pm. Changed the
"executable" class methods in Index.pm, Search.pm and Daemon.pm to use
this class method. Now also returns software and index version
information. Should also return license information in the future
when NexTrieve will also return that on a -V. However, this still
doesn't solve the test-suite errors if NexTrieve is installed but the
license cannot be found or is out of date.
1 February 2002
Started implementation of the MIME-processor concept in RFC822.pm, that
should allow external processors for specific MIME-types to be
specified. Add text/plain and text/x-diff handlers.
Moved "displaycontainers" and "removecontainers" functionality from
HTML.pm to NexTrieve.pm, so that it can be inherited by RFC822.pm.
Changed the "scripts" directory to "script" and added it as "EXE_FILES"
in the Makefile.PL specification. The scripts "docseq",
"mailbox2ntvml" and "html2ntvml" are now automatically installed in
/usr/local/bin if a "make install" is done.
Fixed problem in NexTrieve.pm that would cause test-suite errors if
Text::Iconv was not installed and the Unix "iconv" utility _was_
available.
Added "docseq" script to quickly create a document sequence out of a
bunch of files that were created by another process. Added test for
the script functionality.
Added "files" method to Docseq.pm, to allow for quick merging of pre-
created NTVML-files into a Docseq. Added a special case "read_string"
to Document.pm so that encoding is removedi from read-made XML and
added to the object so that $docseq->files can do its work without
having to create a DOM. Added test for this functionality.
0.04 1 February 2002
Fixed last nit in RFC822.pm which was exposed while testing the
mailbox2ntvml script.
30 January 2002
Ported the NexTrieve standard script "ntvmailbox2ntvml" to use the
new NexTrieve::Mbox module and added it as "mailbox2ntvml" in the
scripts directory.
Completed first version of the NexTrieve::Mbox module + associated
test-suite. You can now easily index one or more standard Unix
mailboxes and have filename, offset and length attributes added
automagically. In concept based on the ntvmailbox2ntvml script in
the NexTrieve distribution.
Added general purpose method "ampersandize" to NexTrieve.pm, as a
subset of what "normalize" does. Changed normalization method of
RFC822 from "normalize" to "ampersandize".
Added Resource method to NexTrieve::RFC822 module. Creates a Resource
object with <indexcreation> section that corresponds to the XML that
is generated by Document.
Changed NexTrieve.pm so that empty containers are always written out
in alphabetical order. This should make the XML more predictable
(as hashes do not have same order in different versions of Perl).
Adapted t/03resource.t to now check again for predictable XML.
Inheritable method "xml" now warns the XML if called in a void context
without any parameters. That mode of operation is intended as a
debugging tool.
Added Resource method to NexTrieve::HTML. Removed the attributes and
texttypes methods in favour if that. Added test to t/12html.t to check
whether it works.
29 January 2002
Completed first version of NexTrieve::RFC822 module. Added support for
extra attributes and texttypes from external sources. Added examples
using this in the test-suite. Internally generalized a lot of stuff,
resulting in less source code at the expense of a little CPU overhead.
Added 'epoch' as a keyed processing routine.
28 January 2002
Nearing completion on the NexTrieve::RFC822 module. Removed the special
"date" type and replaced that by a more generic processing routine
concept. Re-created the date processing as a standard processing
routine named "datestamp", added "timestamp" as an alternate processing
routine that creates timestamp in the form YYYYMMDDHHMMSS.
27 January 2002
Removed test for NexTrievePath from t/01basic.t: it was causing false
failures on platforms where NexTrieve is not installed.
Moved functionality of NexTrieve::HTML->Docseq method to the
NexTrieve.pm module: now any module that inherits from the NexTrieve.pm
only needs to supply a Document() method to be able to create many
NexTrieve::Documents from any data source.
Added support for Text::Iconv to recoding functions of NexTrieve.pm.
Fixed problem in NexTrieve::HTML: removecontainers would only remove
<script> even if other containers were specified.
Started work on the NexTrieve::RFC822 module.
Removed debug nit from NexTrieve::HTML->Docseq that would actually
cause the HTML-file to be converted twice.
0.03 26 January 2002
Adapted NexTrieve's "ntvhtml2ntvml" filter for use with the NexTrieve
module and added as a script named "html2ntvml" and added test of
usage to 12html.t. Adapted MANIFEST accordingly.
Finished first public version of NexTrieve::HTML module and added
test-file 12html.t.
25 January 2002
Fixed up encoding issues over all objects, especially with
NexTrieve::Document and NexTrieve::Docseq. If a document has an
encoding different from the docseq, then the XML will be automatically
converted using the "recode" method in the NexTrieve.pm module.
Added the first automatic recoding handler searching strategy to
method "find_recoding" and added the recoding handler that uses
"iconv".
22 January 2002
Re-arranged the still incomplete NexTrieve::Collection module to
have the major part of its intelligence moved to the new
NexTrieve::Collection::Index module.
Created first version of NexTrieve::Collection::Index module.
21 January 2002
Fixed bug in $deamon->pid: now removes the newline from the string
so that the pid becomes truly numeric.
Started work on NexTrieve::HTML based on the ntvhtml2ntvml script.
Added method "Queries" to NexTrieve::Querylog.
0.02 20 January 2002
$daemon->pid now waits for a max of 5 seconds to see whether the
pid-file appears, before returning with an error.
$daemon->start now returns the object itself: since the return value
of starting the daemon is of little value anyway, it makes more sense
to return the object, so that you can do one-liners.
Fixed problem in $ntv->anyport: older IO::Socket::INET _must_ have a
Listen specification, apparently.
Fixed problem in NexTrieve::Docseq: apparently a string resembling a
a namespace is illegal as an unquoted key value in a hash reference
specification in perl 5.005.
Changed various test from direct comparisons to just checking whether
the object was created without errors: that should teach me not to
depend on the order of keys in a hash.
Fixed problem in NexTrieve.pm with perl 5.005: $object->$method
apparently _must_ be $object->$method();
Fixed problem with $ntv->Search not setting method/value pairs.
Added "command" method to NexTrieve::Replay;
Added "eof" methods to NexTrieve::Querylog and NexTrieve::Replay.
0.01 19 January 2002
First upload to CPAN.
First version for the 2.X generation of NexTrieve. Some code and
concepts were used from the old Nextrieve.pm module (note the
lowercase t) that was written by me in 1995 and heavily used by
all search engines of customers of xxLINK.