12:44 p.m.
Currently the po references contain the file and line number:
#: ../wine-user-en-sgml/wine-user.sgml:82
../wine-user-en-sgml/glossary.sgml:171
What this means is that if someone adds a line near the beginning of
wine-user.sgml we will get thousands of lines of diff the next time
po4a-updatepo is run.
I know po tools can probably hide this issue from users (though that
doesn't help me since I use xemacs) but this is still going to be a
problem when comparing two CVS version of the po file. This is also a
big nuisance when trying to merge CVS branches. Look at the output of
the following command to see what I mean:
cvs diff -r1.4 -r1.5 po/pod/fr.po
Also I feel that references that include the line number are less useful
when translating documentation than when translating the strings used in
a program. The strings in a program are often pretty short, sometimes a
single word like 'Close' and searching for such strings is difficult due
to false positives. Thus for programs the line number information is
quite useful. But, when translating the documentation, each msgid is a
paragraph which usually contains at least a full sentence. So it's easy
to find the relevant place in the source SGML file by searching for a
sequence of two or three words.
So I propose to add a 'porefs' option to po4a-updatepo which would allow
the user to control the addition of references. It would have three
possible values:
* none
No references are added
* noline
References are added but all line numbers are set to '1'. I'm
keeping a line number for compatibility with po tools let the user open
the relevant file by double clicking on the reference.
* full
This would be the default and work just as it does today.
Now, the logical place to implement this is the Po.pm module but it was
not possible to pass options to it. So I added a -O option which passes
options to the Po.pm module (via TransTractor,pm) instead of the format
module.
Let me know if this sounds ok or if another approach should be used.
Changelog:
* po4a-updatepo
lib/Locale/Po4a/TransTractor.pm
lib/Locale/Po4a/Po.pm
Francois Gouget <fgouget(a)codeweavers.com>
Add a porefs option to let the user customize the generation of po
references.
Add a -O option to po4a-updatepo so users can pass options to the
Po,pm module.
--
Francois Gouget
fgouget(a)codeweavers.com
Index: po4a-updatepo
===================================================================
RCS file: /cvsroot/po4a/po4a/po4a-updatepo,v
retrieving revision 1.36
diff -u -p -r1.36 po4a-updatepo
--- po4a-updatepo 30 May 2005 07:13:38 -0000 1.36
+++ po4a-updatepo 28 Jun 2005 16:48:11 -0000
@@ -136,7 +136,7 @@ Getopt::Long::config('bundling', 'no_get
# Parse our options
my (@masterfiles,@pofiles);
-my ($help,$help_fmt,$verbose,$debug,$format,@options);
+my ($help,$help_fmt,$verbose,$debug,$format,@options,@po_options);
my $mastchar;
GetOptions('help|h' => \$help,
'help-format' => \$help_fmt,
@@ -148,6 +148,7 @@ GetOptions('help|h' => \$help,
'master-charset|M=s' => \$mastchar,
'option|o=s' => \@options,
+ 'O=s' => \@po_options,
'verbose|v' => \$verbose,
'debug|d' => \$debug,
@@ -169,6 +170,16 @@ foreach (@options) {
$options{$_}=1;
}
}
+
+my %po_options;
+foreach (@po_options) {
+ if (m/^([^=]*)=(.*)$/) {
+ $po_options{$1}="$2";
+ } else {
+ $po_options{$_}=1;
+ }
+}
+$options{'po_options'}=\%po_options;
# parser
my ($doc)=Locale::Po4a::Chooser::new($format,%options);
Index: lib/Locale/Po4a/TransTractor.pm
===================================================================
RCS file: /cvsroot/po4a/po4a/lib/Locale/Po4a/TransTractor.pm,v
retrieving revision 1.67
diff -u -p -r1.67 TransTractor.pm
--- lib/Locale/Po4a/TransTractor.pm 30 May 2005 07:05:23 -0000 1.67
+++ lib/Locale/Po4a/TransTractor.pm 28 Jun 2005 16:48:13 -0000
@@ -330,7 +331,8 @@ sub new {
# private data
$self->{TT}=();
$self->{TT}{po_in}=Locale::Po4a::Po->new();
- $self->{TT}{po_out}=Locale::Po4a::Po->new();
+ $self->{TT}{po_out}=Locale::Po4a::Po->new($options{'po_options'});
+ delete $options{'po_options'};
# Warning, this is an array of array:
# The document is splited on lines, and for each
# [0] is the line content, [1] is the reference [2] the type
Index: lib/Locale/Po4a/Po.pm
===================================================================
RCS file: /cvsroot/po4a/po4a/lib/Locale/Po4a/Po.pm,v
retrieving revision 1.43
diff -u -p -r1.43 Po.pm
--- lib/Locale/Po4a/Po.pm 30 May 2005 07:15:21 -0000 1.43
+++ lib/Locale/Po4a/Po.pm 28 Jun 2005 16:48:12 -0000
@@ -46,6 +46,16 @@ translate everything, including document
package description, debconf templates, and everything which may benefit
from this.
+=head1 OPTIONS ACCEPTED BY THIS MODULE
+
+=over 4
+
+=item porefs
+
+This specifies the reference format. It can be one of 'none' to not produce
+any reference, 'noline' to not specify the line number, and 'full' to
+include complete references.
+
=cut
@@ -93,11 +103,11 @@ a po file we should load.
=cut
sub new {
- my $this = shift;
+ my ($this, $options) = (shift, shift);
my $class = ref($this) || $this;
my $self = {};
bless $self, $class;
- $self->initialize();
+ $self->initialize($options);
my $filename = shift;
$self->read($filename) if defined($filename) && length($filename);
@@ -105,10 +115,18 @@ sub new {
}
sub initialize {
- my $self = shift;
+ my ($self, $options) = (shift, shift);
my $date = `date +'%Y-%m-%d %k:%M%z'`;
chomp $date;
+ $self->{options}{'porefs'}= 'full';
+ foreach my $opt (keys %$options) {
+ if ($options->{$opt}) {
+ die wrap_mod("po4a::po", dgettext ("po4a", "Unknown option:
%s"), $opt) unless exists $self->{options}{$opt};
+ $self->{options}{$opt} = $options->{$opt};
+ }
+ }
+
$self->{po}=();
$self->{count}=0;
$self->{header_comment}=
@@ -641,7 +659,7 @@ sub gettext {
my $res;
return "" unless defined($text) && length($text); # Avoid returning
the header.
- my $validoption="wrap wrapcol";
+ my $validoption="reference wrap wrapcol";
my %validoption;
map { $validoption{$_}=1 } (split(/ /,$validoption));
@@ -850,7 +868,13 @@ sub push_raw {
$self->{header_comment}=$comment;
return;
}
-
+
+ if ($self->{options}{'porefs'} eq "none") {
+ $reference = "";
+ } elsif ($self->{options}{'porefs'} eq "noline") {
+ $reference =~ s/:[0-9]*/:1/g;
+ }
+
if (defined($self->{po}{$msgid})) {
warn wrap_mod("po4a::po", dgettext("po4a","msgid defined
twice: %s"), $msgid) if (0); # FIXME: put a verbose stuff
if ($msgstr && $self->{po}{$msgid}{'msgstr'}
2:50 p.m.
On Tue, Jun 28, 2005 at 07:44:23PM +0200, Francois Gouget wrote:
Currently the po references contain the file and line number:
#: ../wine-user-en-sgml/wine-user.sgml:82
../wine-user-en-sgml/glossary.sgml:171
What this means is that if someone adds a line near the beginning of
wine-user.sgml we will get thousands of lines of diff the next time
po4a-updatepo is run.
I know po tools can probably hide this issue from users (though that
doesn't help me since I use xemacs) but this is still going to be a
problem when comparing two CVS version of the po file. This is also a
big nuisance when trying to merge CVS branches. Look at the output of
the following command to see what I mean:
cvs diff -r1.4 -r1.5 po/pod/fr.po
Also I feel that references that include the line number are less useful
when translating documentation than when translating the strings used in
a program. The strings in a program are often pretty short, sometimes a
single word like 'Close' and searching for such strings is difficult due
to false positives. Thus for programs the line number information is
quite useful. But, when translating the documentation, each msgid is a
paragraph which usually contains at least a full sentence. So it's easy
to find the relevant place in the source SGML file by searching for a
sequence of two or three words.
I always play with -I^#: but I agree that this may be suboptimal, and with
your point in general.
So I propose to add a 'porefs' option to po4a-updatepo which
would allow
the user to control the addition of references. It would have three
possible values:
* none
No references are added
* noline
References are added but all line numbers are set to '1'. I'm
keeping a line number for compatibility with po tools let the user open
the relevant file by double clicking on the reference.
* full
This would be the default and work just as it does today.
Now, the logical place to implement this is the Po.pm module but it was
not possible to pass options to it. So I added a -O option which passes
options to the Po.pm module (via TransTractor,pm) instead of the format
module.
I don't really like it. Users shouldn't really care about whether the option
is for this part of the code or that one. Couldn't we seed the option set of
each modules from the transtractor code or such?
Let me know if this sounds ok or if another approach should be used.
I was about commiting this as is, and I then thought that I'll never have
the courage to do it right if I don't do so right now.
Mixing module options and "general" ones (ie, for Po.pm or so) would also
allow to add this feature to all binaries, not only updatepo. The price to
pay should only be to come up with a common way to handle options accross
modules and binaries (unless I missed something). That sounds doable.
I'll try to give it a spin tomorow.
Mt.
6:15 p.m.
Martin Quinson wrote:
[...]
I always play with -I^#: but I agree that this may be suboptimal, and
with
your point in general.
Hmm, what' -I^#?
[...]
>Now, the logical place to implement this is the Po.pm module but
it was
>not possible to pass options to it. So I added a -O option which passes
>options to the Po.pm module (via TransTractor,pm) instead of the format
>module.
I don't really like it. Users shouldn't really care about whether the option
is for this part of the code or that one.
Agreed.
Couldn't we seed the option set of
each modules from the transtractor code or such?
I've tried something like this but it failed because Sgml.pm complained
it did not know about 'porefs'.
Maybe the modules should add the options they recognize to a hash list
that they would return to whoever controls things at the top. Then any
option missing from the hash would be reported as an error.
Mixing module options and "general" ones (ie, for Po.pm or
so) would also
allow to add this feature to all binaries, not only updatepo.
Yes, po4a-gettextize would probably need this option too.
[...]
I'll try to give it a spin tomorow.
Great!
--
Francois Gouget
fgouget(a)codeweavers.com
6:45 p.m.
On Wed, Jun 29, 2005 at 01:15:28AM +0200, Francois Gouget wrote:
Martin Quinson wrote:
[...]
>I always play with -I^#: but I agree that this may be suboptimal, and with
>your point in general.
Hmm, what' -I^#?
How to instruct diff to ignore ignorable changes
[...]
>>Now, the logical place to implement this is the Po.pm module but it was
>>not possible to pass options to it. So I added a -O option which passes
>>options to the Po.pm module (via TransTractor,pm) instead of the format
>>module.
>
>I don't really like it. Users shouldn't really care about whether the
>option
>is for this part of the code or that one.
Agreed.
> Couldn't we seed the option set of
>each modules from the transtractor code or such?
I've tried something like this but it failed because Sgml.pm complained
it did not know about 'porefs'.
Maybe the modules should add the options they recognize to a hash list
that they would return to whoever controls things at the top. Then any
option missing from the hash would be reported as an error.
It was somehow my idea. Will see if it works...
>Mixing module options and "general" ones (ie, for Po.pm
or so) would also
>allow to add this feature to all binaries, not only updatepo.
Yes, po4a-gettextize would probably need this option too.
[...]
>I'll try to give it a spin tomorow.
Great!
To be continued.
Mt.
2:08 a.m.
On Wed, Jun 29, 2005 at 01:15:28AM +0200, Francois Gouget wrote:
Martin Quinson wrote:
> Couldn't we seed the option set of
>each modules from the transtractor code or such?
I've tried something like this but it failed because Sgml.pm complained
it did not know about 'porefs'.
Maybe the modules should add the options they recognize to a hash list
that they would return to whoever controls things at the top. Then any
option missing from the hash would be reported as an error.
Now, Transtractor adds "porefs" to the $self->{options} hash before calling
the initialization of the plugin, so they won't complain about it being
unknown. I checked, all the plugins accepting options deal with them the
same way in their initialize().
Then, Transtractor uses its value to build the Po->new() options.
I guess it's good enough for now, but would appreciate feedback on this one
too.
Bye, Mt.
6851
days inactive
6876
days old
4 comments
2 participants
participants (2)
-
Francois Gouget -
Martin Quinson