$Id: Changes,v 1.75 2005/07/27 12:32:53 mike Exp $
Revision history for Perl extension Net::Z3950.
0.50 Wed Jul 27 13:28:03 BST 2005
- Correct the ResultSet::present() fix that was supposed to be
in the previous release: that code inadvertently always
returned undef in asynchronous mode.
- "ywpriv.h" now #undefines "list", which the Perl development
header files inexplicably and inexcusably #define to
"Perl_list".
- "ywpriv.h" now #undefines "open", which Solaris 9 #defines
(I really can't believe this) to "open64". Maybe Sun would
like me to go round their place and dump a load of _my_ crap
in _their_ workspace?
- New querytype "cql" can be used to pass CQL queries,
untranslated, straight through to the server. Use like:
$conn->search(cql => "au=(kernighan and richie)")
0.49 Thu Apr 21 11:06:02 BST 2005
- Change to the ResultSet::present() API in asynchronous mode:
returns 1 if new requests were queued, 0 if all of the
requested records had already been queued. (No changes to
its usage in synchronous mode.)
- yazwrap/send.c no longer includes <yaz/log.h> (not needed,
generates warning)
0.48 Tue Apr 19 22:36:45 2005
- Patches supplied by Jan Bauer Nielsen <jbn@dtv.dk> to
interpret otherInformation packets, in search responses,
containing a search-info report of subquery counts.
- Patches supplied by Myron Turner <mturner@Ms.UManitoba.CA>
to protect the Net::Z3950 module against certain classes of
badly-behaved server.
- Bring "samples/multiplex.pl" up to date, illustrating the
use of this module to run sequences of queries against
multiple servers in parallel.
0.47 Tue Jan 4 22:15:07 2005
- The "connection refused" condition is now handled more
gracefully, so that there is no error message emitted by
Event.pm, just an undefined value returned and $! set to
ECONNREFUSED.
- Improve many of the comments on this subject.
0.46 Mon Nov 22 23:42:40 2004
- Use the MARC::Record module for rendering MARC records
instead of the older, unsupported and possibly buggy
MARC.pm.
WARNING: This change is backwards incompatible, but that
incompatibility will only affect you if you're relying on
the precise formatting of $rec->render() on MARC records.
0.45 Mon Nov 1 09:13:51 2004
- Support for "charset" and "language" options added by Adam
Dickmeiss.
0.44 Fri May 7 18:00:54 2004
Note that this release consists entirely of
backwards-incompatible changes to new functionality introduced
in 0.43, the previous release. If you use Scan, then all your
Scan code must change; if you do not use Scan, then this
release will not affect you at all.
- Change scan-option names to be ZOOM-compliant.
- Introduce a new enumeration, Net::Z3950::ScanStatus
- Add a new class, Net::Z3950::ScanSet, representing the
results of a Scan operation. This is compliant with the
ZOOM Abstract API, and supports much nicer client code than
the previous release:
$ss = $conn->scan('@attr 1=1003 kern');
$status = $ss->status();
print $ss->errmsg()
if $status == Net::Z3950::ScanSet::Failure;
$count = $ss->size();
($term, $hits) = $ss->term(0);
The scanResponse APDU, which used to be returned from the
scan() method, is still available via the scanResponse()
method, but there is no reason for new code to call this.
- Complete rewrite of the sample client "scan.pl" to use the
new API.
0.43 Thu May 6 14:19:56 2004
- Add support for Scan, using code contributed by Jan Bauer
Nielsen <jbn@dtv.dk>, somewhat tidied up. Use it like this:
$apdu = $conn->scan('@attr 1=1003 kern');
$status = $apdu->scanStatus();
if ($status == 6) print Dumper($apdu->diag());
$count = $apdu->numberOfEntriesReturned();
$term = $apdu->entries()->[0]->termInfo()->term()->general();
$hits = $apdu->entries()->[0]->termInfo()->globalOccurrences();
Subsequent releases will support more corners of the Scan
specification (e.g. displayTerm and friends), and present a
much nicer API with a ZOOM-like ScanSet object. See
http://zoom.z3950.org/api/zoom-1.4.html#3.6
- Included in the "samples" directory is a simple scanning
client program, "scan.pl".
0.42 Wed Mar 31 13:06:09 2004
- Add support for Unix-domain sockets: use "unix" as the
hostname, and the path to the socket file as the port, in
the connection constructor. Like this:
$conn = new Net::Z3950::Connection("unix", "/tmp/socket");
0.41 Wed Mar 17 13:33:43 2004
- When Net::Z3950::Manager::wait() times out, it now returns
an undefined value rather than, as in v0.40, the timeout
value. This code is now tested, thanks to Rick Jansen.
- Minor changes to some of the logic in ResultSet.pm
0.40 Tue Mar 16 13:57:33 2004
- A timeout may be set for a Net::Z3950::Manager's wait() by
setting that manager's "timeout" option to the maximum
number of seconds to wait. Donald Knuth once wrote,
"Beware: this code has not been tested, only proved
correct." _This_ code improves on Knuth in that it has been
neither tested _nor_ proved correct. However, I give you my
solemn promise that I think it probably works, as likely as
not.
0.39 Fri Dec 19 16:08:35 2003
- Recognise record-syntax names such as "USMARC" as well as
enumerators such as Net::Z3950::RecordSyntax::USMARC, in
accordance with what's specified in v1.4 of the ZOOM AAPI.
Such names are recgonised case- and hyphen-insensitively, so
"GRS-1" and "grs1" are equivalent.
- Add "XML" as an alias for the "TEXT_XML" record-syntax.
- Handle the ZOOM AAPI's standard options "pass" and "group"
in preference to the old options "password" and "groupid"
(although these are still also recognised, for the benefit
of old applications.)
- Handle the ZOOM AAPI's standard option "async" in preference
to the old "type" with _value_ "async" (although that's
still also recognised, for the benefit of old applications.)
- Makefile.PL includes support for and instructions about
building against a YAZ in a non-standard place ("non-root
install").
- Another slight refinement to samples/simple.pl's
diagnostics.
- Documentation for diagnostic methods now includes a link to
the BIB-1 diagnostics web-page.
0.38 Fri Oct 24 13:28:42 2003
- Fix a silly bug in Net::Z3950::Record::OPAC::rawdata(),
spotted by Kristina Long. It was trying to return $$this,
like the opaque-array-of-bytes record-types such as USMARC,
whereas in fact $this is itself a Perl structure
representing the record.
- Add a kludge to patch over anti-social behaviour from
servers that sometimes return USMARC records when asked for
OPAC records: the returned record is wrapped in an otherwise
empty OPAC record structure.
0.37 Tue Sep 16 15:14:40 2003
- Remove a pair of extraneous declarations (copied from
"yaz/include/yaz/z-opac.h" for reference) at the end of
translateHoldingsAndCirc(). On recent GCC versions, such as
mine, they were pointless but harmless; but older versions,
such as the one in Debian/stable, croak if you mix
statements and declarations.
- Improve behaviour when a server unilaterally closes the
connection while we're waiting for a Init response.
Previously, this would cause a segmentation fault; now you
get errno set to 104 (ECONNRESET) which is pretty close.
- Improve the OPAC record renderer in Z3950/Record.pm so that
it walks more of the OPAC-record structure. As well as
being prettier, this is better documentation-by-example of
how to use the record structure.
0.36 Fri Sep 12 23:47:00 2003
- Add support for the OPAC record syntax: OPAC records are now
translated into self-describing Perl data-structures,
blessed into the class Net::Z3950::Record::OPAC. There is a
rudimentary renderer included in that class, but
sophisticated applications will want to deal with the data
structure itself. To see it in action, use something like:
$ perl samples/simple.pl alpha.calarts.edu 210 \
MARION fruit preferredRecordSyntax 35
WARNING: for some servers, e.g. troy.lib.sfu.ca:210/innopac,
you MUST request elementSetName "F", otherwise you'll just
be fed USMARC records whatever your preferredRecordSyntax.
- Deal better with unrecognised or malformed APDUs from the
server, e.g. library.ucfv.bc.ca:2200/unicorn when asked
for the USMARC record found by "@attr 1=4 time". Now, the
operation fails with diagnostic code 1001 (Malformed APDU).
- Change all <mike@tecc.co.uk> addresses to
<mike@indexdata.com>. (This change is well overdue, since I
changed jobs seven months ago!)
- Fix reference for PQF in the YAZ manual.
0.35 Fri Jun 27 10:48:03 2003
- Fix all non-US MARC record-types to return raw data from the
rawdata() method rather than, as before, the render()
method.
- Add a new delete() method to Net::Z3950::ResultSet
0.34 Mon May 12 10:17:07 2003
- Slightly more robust cleanup in Manager::forget()
- Make samples/simple.pl a little less simple(!) by making a
manager so that pre-connection options such as
authentication parameters can be set.
- In Makefile.PL, LIBS and INC are now both set via yaz-config,
so you don't need to mess with that file before building.
(Adam Dickmeiss)
- The INC is now inherited by yazwrap/Makefile.PL, so that
this file also does not need to be edited. (Adam Dickmeiss)
- CCFLAGS is no longer set in the yazwrap directory. I only
did that in order to add "-Wall", which is pleasant luxury
but not really necessary; and doing so caused problems as
other, default, components of that setting were discarded on
some platforms. (Adam Dickmeiss)
- Add support for the MAB record syntax, thanks to Heiko
Jansen <JANSEN@hbz-nrw.de>
0.33 Thu Apr 3 09:10:16 2003
- Add explicit statement of GPL licence.
No code changes at all.
0.32 Tue Jan 21 16:46:23 2003
- Remove the old, unnecessary and confusing code that patched
around the absence of cs_look() and cs_rcvconnect() in YAZ.
In fact, they've both been there since version 1.8 (the one
that introduced ZOOM-C), fourteen months ago. This change
cleans up the code and fixes an intermittent error that Dave
Mitchell had run into. You now need v1.8 of YAZ or newer.
0.31 Wed Nov 27 12:40:50 2002
- Add command-line options to samples/simple.pl
- Add samples/canonical.pl
- Change the default element-set names in Manager.pm to be
upper-case "F" and "B" rather than "f" and "b", as specified
in section 3.6.2 (Composition Specification, Comp-spec
Omitted) of the standard.
http://lcweb.loc.gov/z3950/agency/markup/08.html#3.6.2
Even though a clarification makes it clear that element-set
names should be treated case-insensitively.
http://lcweb.loc.gov/z3950/agency/clarify/esncase.html
- Formally deprecate the non-working records() method of the
ResultSet class.
- When a new Connection object can't be made because of Init
being refused (rather than due to a lower-level error such
as ECONNREFUSED), $! is set to the distinguished value -1.
(Dave Mitchell)
- Neater code for connection-closing.
(Dave Mitchell)
- Fake up a diagnostic record (BIB-1 error 3, "unsupported
search") for servers which illegally fail to include
diagnostic records in the result of a failed search.
(Dave Mitchell)
- Added #include "XSUB.h" to the yazwrap directory's private
code, which means it now does The Right Thing when compiling
against a Perl build with the MULTIPLICITY flag set. This
means that we now build and run under cygwin on Windows 2000
(Dave Mitchell)
- Add new option "namedResultSets", initially true, which if
set false caters for brain-damaged servers by always using
the same result-set name, "default", thereby invalidating
previous ResultSet objects.
(Dave Mitchell)
- Documentation tweaks.
0.30 Fri Jul 19 13:55:45 2002
Today's release is brought to you entirely by Dave Mitchell
<davem@fdgroup.com>, who has done an astonishing amount of
really useful work on Net::Z3950. Thanks, Dave!
- Added the present() method and "prefetch" option to optimise
multiple calls of record() in synchronous mode.
- Made $conn->close() actually clean up and close the connection
- Fixed memory leak in receive.c, decodeAPDU()
- Added functions to decode close request/response APDUs.
- Really do add ability to set custom exception-handler via
"die_handler" option -- I [Mike] did it wrong last time.
- Changed all print()s to die()s, which can be caught as
exceptions.
- Commented out a debugging warn()
- Added a dummy "test" target to yazwrap/Makefile.PL to stop
it complaining during "make test" [At last! -- Mike]
- Removed "debug => 5" from watcher creators.
- Updated the synchronous synopsis:
* Replaced use of records() with record(), since the former
is deprecated.
* Added error-checking, so that people who cut-and-paste the
synposis start off with good habits.
* Added $conn->close(), so people know this doesn't happen
automatically.
0.29 Fri May 17 17:07:53 2002
- Add ability to set custom exception-handler via
'die_handler' option.
- Fix dereference-undef bug occurring when performing a second
search on a connection without having fetched any records
following the first search.
- Cope with servers which lie about the number of records in
their present responses (the cads!)
- Fix to correctly translate OIDs in which one or more
components is "0".
All of these changes are due to Dave Mitchell
<davem@fdgroup.com> -- thanks, Dave!
0.28 Wed Feb 27 17:28:26 2002
- Change representation of OIDs from blessed reference to
array of intergers to Boring Old String (dot-separated
numbers). Resulting tweaks elsewhere in the module, e.g. in
GRS::render()
- Fix Tutorial's incorrect description of the GRS-1 record
structure. How come no-one noticed this gross error in all
the time it was there?
- Minor fixes to the APDU class's documentation (but you
shouldn't be reading that anyway :-)
- Fix the top-level documentation file (index.html) to use the
same stylesheet as all the API documentation.
0.27 Mon Feb 11 12:59:05 2002
- Incorporate a patch from Adam <adam@indexdata.dk> to allow
Net::Z3950 to work against servers that don't support named
result sets. Ah, so simple, so elegant!
0.26 Fri Feb 8 16:06:15 2002
- OK. This time I have _really_ fixed connecting to work
properly either synchronously or asynchronously, and
slightly clarified some of the event-handling code while I
was at it. In particular, ECONNREFUSED is handled properly
now.
- Add AUTHOR and ABSTRACT to my Makefile.PL, as recommended in
passing by R. Geoffrey Avery's Lightning Talk _h2xs has
eXceeded its Stay_, at
http://www.platypiventures.com/perl/present/tpc5modules/004.html
0.25 Tue Jan 29 15:56:48 2002
- Remove some obsolete comments (documenting bugs that are now
fixed!)
- Fix type mismatch in encode_APDU() -- we've been getting
away with it until now because sizeof(size_t)==sizeof(int)
on most systems, but not on the 64-bit DEC Alpha OSF/1.
- Fix daft-but-harmless bug in yazwrap/receive.c's setString()
functions, in which we return the value of a void function.
Most compilers DWIM, but not all.
- Add explicit casts to (char*) in yazwrap/receive.c where I'd
been assuming the conversion from unsigned char*. Ho hum.
0.24 Wed Jan 23 11:39:32 2002
- Fix a truly stupid bug which prevents synchronous connection
from working. I missed this in 0.23 because the test script
-- unlike almost every other Net::Z3950 program -- uses
asynchronous connection. D'oh!
0.23 Tue Jan 22 16:11:41 2002
- Handle surrogate diagnostics (e.g. bad element set name).
- Finally fix that stupid "No rule to make target `pure_all'"
message that "make test" always used to produce in the
yazwrap directory.
- Add support for retrieving HTML records: new record-syntax
enumeration TEXT_HTML, and a new record-type class
Net::Z3950::Record::HTML.
- Fix hanging on ECONNREFUSED.
0.22 Fri Oct 19 16:37:44 2001
- Add rudimentary ("straight through") support for the OPAC
record syntax.
- Fix URLs for perl.z3950.org to omit obsolete /nz/ directory.
- Fix "make clean" in doc directory to remote the Z3950
subdirectory.
- Fix documentation stylesheet to match perl.z3950.org
0.21 Thu Oct 18 14:29:06 2001
- Fix samples/multiplex.pl not to die when one server fails a
search - it just prints and error message now.
- Get rid of stupid "field `DESTROY' not defined" messages.
0.20 Fri Oct 12 17:12:04 2001
- Add name() method to the Net::Z3950::Connection class.
- Add the wherewithal for asynchronous operations to invoke
callbacks when they complete
- Fix staggeringly stupid bug in which the synchronous
search() method would always expect an initResponse, so that
you couldn't call search() twice on one connection.
- Add support for the UNIMARC record syntax (required by Bath)
- Remove the stupid and implemented records() method from the
Connection class.
- Implement CCL qualifiers, specified in "ccl.qual" file.
- Improve diagnostics from APDU assembly.
- Fix a rather silly bug where records, once fetched, were
cached irrespective of their element-set name, so that
record fetched as 'b' and then as 'f' would appear brief.
- Revamp the "samples" directory: all examples now work(!)
- Add ZOOM references to the documentation.
- Minor fixes to documentation, comments, etc.
(This major revision - indicated by the increment of the
first-decimal-place version number - was largely motivated
by the October 2001 ZIG meeting's tutorial, for which I
prepared a simple multiplexing client, and was surprised at
how hard it turned out to be.)
0.14 Wed Jul 18 13:25:06 2001
- Add support for retrieving XML records: new record-syntax
enumerations TEXT_XML and APPLICATION_XML, and a new
record-type class Net::Z3950::Record::XML.
- Minor fixes to documentation, comments, etc.
0.13 Fri Jun 22 09:30:55 2001
- In yazwrap/util.c, change socklen_t to size_t since the
former does not seem to be defined on all platforms. In
particular, Solaris 2.6 uses a size_t* for the last argument
to getsockopt().
0.12 Wed Feb 21 16:47:03 2001
- Change my email address and web-page in the README.
- Add conditionally-compiled debugging code to Yaz-wrapper.
- Add rudimentary support for rendering MARC records and
remove the MARC-related apology from the tutorial.
- Add note to README about "make test" error message.
- Add note to README about support options, including the
Net::Z3950 mailing list.
0.11 Fri Feb 9 11:31:34 2001
- Fix Net::Z3950::ResultSet->record() to work correctly in
synchronous mode (i.e. wait for the record to become
available before returning.)
- Add rawdata() method to the various Net::Z3950::Record::*
classes: we now have a distinction between render() and
rawdata(), concepts which were previously mulched in
together very unhelpfully.
- Add Tutorial.
- FIRST PUBLIC RELEASE ON CPAN
0.10 Wed Nov 29 16:39:47 2000
- Change module name from Z3950 to Net::Z3950
0.04 Fri Sep 22 17:51:28 2000
- Fix so that diagnostics are available after search failure.
- Add trivial "straight through" support for USMARC and some
of the related record syntaxes.
- Support for OID-typed fields in GRS-1 records.
- Change default database to "Default" (it had the profoundly
unhelpful default of "hardwired-databaseName"!)
- Modify the internal class hierarchy to make MARC support
hang together more easily (yazwrap library now returns
...::Record::<RS> objects instead of ...::APDU::<RS>)
- Improve diagnostics when low-level translation of C objects
into Perl objects fails.
0.03 Thu Jun 29 15:38:33 2000
- First version to be held in CVS at indexdata.dk
- Add CVS identification headers where they're missing, fix
some that lacked the trailing $-sign (duh.)
- Fix the search() and records() wrapper methods so that much
simpler single-threaded clients can be easily written.
0.02 Tue Jun 27 21:28:56 2000
- Make the $port argument optional in the Z3950::Connection
constructor.
- Add support for mnchar* (T_MNPV) to the typemap file: this
is a Maybe-Null PV, into which undefined values may be
passed yielding null pointers.
- Add samples/fetch1.pl, which actually works (unlike the
other scripts in the samples directory, which are written to
a pre-0.1 version of the interface.)
- Remove obsolete bits of the README file.
0.01 Tue May 23 09:20:30 2000
- Originally created by h2xs 1.19
- 0.01 was the first numbered version. Earlier versions were
never assigned numbers as they never got outside the
author's house; this is the first one that was sent to Index
Data.
--
To be done:
- ### Fix so that when one of the Connection objects
associated with an asynchronous Manager can't connect,
$mgr->wait() returns a reference to that Connection, and
arranges for $conn->op() to return Net::Z3950::Op::Error,
and leaves things in a state whereby wait() can be
successfully re-invoked to continue the concurrent
operations of the remaining Connections. This is
potentially tricky to implement and test, and will only be
used by a small proportion of applications, so I am not
letting it delay release of other new functionality.
- ### Fix test.pl to work against either local or remote server.
- ### Make managers drop references to closed connections.
- ### Make $mgr->wait() finish when there are no connections left.
- ### Do away with the "async" option, probably. Here's what
it actually does when turned on (the default):
- Connection::new() waits for an Init response before
returning.
- ResultSet::present() (and hence ResultSet::record())
waits for for a Present response before returning.
See also the comment on the Connection::search()
method.