man2po2tmx2po2man \n bug

16.8.2008 17:47 | Přečteno: 980× | man-pages-cs | poslední úprava: 16.8.2008 17:50

Tento zápisek slouží pouze jako podklad pro bugreport. Navíc je anglicky.
This entry is just an attachment to a bugreport. Nothing interesting here.

My goal is to move the current Linux manual pages translation framework (translating to czech) to OmegaT. As OmegaT handles .properties and .po files only, I'm using po4a to transform the man pages to po files and back. To ease the transition I'm creating translation memories from existing translations using po2tmx.

The original en/intro.1 looks like this:

...
.BI "knuth login: " aeb
.BI "Password: " ********
.BI "% " date
Tue Aug  6 23:50:44 CEST 2002
.BI "% " cal
     August 2002
Su Mo Tu We Th Fr Sa
             1  2  3
 4  5  6  7  8  9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31

.BI "% " ls
...

Its czech translation cs/intro.1 looks like this:

...
.BI "knuth login: " aeb
.BI "Password: " ********
.BI "% " date
Út Srp  6 23:50:44 CEST 2002
.BI "% " cal
     Srpen 2002
Ne Po Út St Čt Pá So
             1  2  3
 4  5  6  7  8  9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31

.BI "% " ls
...

This is how it looks when viewed using man en/intro.1 or man cs/intro.1:

To create a po file from the man pages I use the following command:

$ po4a-gettextize -f man -m en/intro.1 -l cs/intro.1 -L utf8 -p intro.1.po

The relevant part of the created intro.1.po file looks like this:

...
# type: Plain text
#: en/intro.1:106
#, fuzzy, no-wrap
msgid ""
"B<knuth login: >I<aeb>\n"
"B<Password: >I<********>\n"
"B<% >I<date>\n"
"Tue Aug  6 23:50:44 CEST 2002\n"
"B<% >I<cal>\n"
"     August 2002\n"
"Su Mo Tu We Th Fr Sa\n"
"             1  2  3\n"
" 4  5  6  7  8  9 10\n"
"11 12 13 14 15 16 17\n"
"18 19 20 21 22 23 24\n"
"25 26 27 28 29 30 31\n"
msgstr ""
"B<knuth login: >I<aeb>\n"
"B<Password: >I<********>\n"
"B<% >I<date>\n"
"Út Srp  6 23:50:44 CEST 2002\n"
"B<% >I<cal>\n"
"     Srpen 2002\n"
"Ne Po Út St Čt Pá So\n"
"             1  2  3\n"
" 4  5  6  7  8  9 10\n"
"11 12 13 14 15 16 17\n"
"18 19 20 21 22 23 24\n"
"25 26 27 28 29 30 31\n"
...

I have to remove the ', fuzzy' indications that were automatically inserted by po4a, otherwise po2tmx would ignore all the messages:

$ sed -i 's/, fuzzy//' intro.1.po

Because of bug 2021007 in OmegaT, I also have to remove the strings ', no-wrap' from the po file:

$ sed -i 's/, no-wrap//' intro.1.po

Then I create a tmx file from the po file:

$ po2tmx -l cs -i intro.1.po -o intro.1.tmx

The relevant part of the intro.1.tmx file looks like this:

...
                <tu>
                        <tuv xml:lang="en">
                                <seg>B&lt;knuth login: &gt;I&lt;aeb&gt;
B&lt;Password: &gt;I&lt;********&gt;
B&lt;% &gt;I&lt;date&gt;
Tue Aug  6 23:50:44 CEST 2002
B&lt;% &gt;I&lt;cal&gt;
     August 2002
Su Mo Tu We Th Fr Sa
             1  2  3
 4  5  6  7  8  9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31
</seg>
                        </tuv>
                        <tuv xml:lang="cs">
                                <seg>B&lt;knuth login: &gt;I&lt;aeb&gt;
B&lt;Password: &gt;I&lt;********&gt;
B&lt;% &gt;I&lt;date&gt;
Út Srp  6 23:50:44 CEST 2002
B&lt;% &gt;I&lt;cal&gt;
     Srpen 2002
Ne Po Út St Čt Pá So
             1  2  3
 4  5  6  7  8  9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31
</seg>
                        </tuv>
                </tu>
...

Then I launch OmegaT, select 'Project'->'New' and give 'manpagestest' as the new project's name. The Source files language is en-us, target language is cs. Sentence-level segmenting is disabled. I quit the project.
Then I copy the tmx to the project:

$ cp intro.1.tmx manpagestest/tm/

Then I create an english po file from the original man page and place it to the source directory of the project:

$ po4a-gettextize -f man -m en/intro.1 -p manpagestest/source/intro.1.po

Then I launch OmegaT, open the manpagestest project and click on the intro.1.po file to edit it. When I change to the message containing the calendar, I can see that the suggested translation (based on the tmx) is not 100% as with other strings, it's only 67 % :

Still hoping it will turn out ok I put the fuzzy match as new translation:

save the project and click 'Project'->'Create translated documents'. I close OmegaT now.

The relevant part of the translated manpagestest/target/intro.1.po file now looks like this:

...
# type: Plain text
#: en/intro.1:106
#, no-wrap
msgid ""
"B<knuth login: >I<aeb>\n"
"B<Password: >I<********>\n"
"B<% >I<date>\n"
"Tue Aug  6 23:50:44 CEST 2002\n"
"B<% >I<cal>\n"
"     August 2002\n"
"Su Mo Tu We Th Fr Sa\n"
"             1  2  3\n"
" 4  5  6  7  8  9 10\n"
"11 12 13 14 15 16 17\n"
"18 19 20 21 22 23 24\n"
"25 26 27 28 29 30 31\n"
msgstr B<knuth login: >I<aeb>
B<Password: >I<********>
B<% >I<date>
Út Srp  6 23:50:44 CEST 2002
B<% >I<cal>
     Srpen 2002
Ne Po Út St Čt Pá So
             1  2  3
 4  5  6  7  8  9 10
11 12 13 14 15 16 17
18 19 20 21 22 23 24
25 26 27 28 29 30 31
...

Obviously this is already wrong but for the sake of the experiment, I'm going on:

$ po4a-translate -f man -m en/intro.1 -p manpagestest/target/intro.1.po -l intro.cs.1
...
manpagestest/target/intro.1.po:134: (po4a::po)
Strange line: --><--
manpagestest/target/intro.1.po:165: (po4a::po)
Strange line: -->msgstr B<knuth login: >I<aeb><--
manpagestest/target/intro.1.po:166: (po4a::po)
Strange line: -->B<Password: >I<********><--
manpagestest/target/intro.1.po:167: (po4a::po)
Strange line: -->B<% >I<date><--
manpagestest/target/intro.1.po:168: (po4a::po)
Strange line: -->Út Srp  6 23:50:44 CEST 2002<--
manpagestest/target/intro.1.po:169: (po4a::po)
Strange line: -->B<% >I<cal><--
manpagestest/target/intro.1.po:170: (po4a::po)
Strange line: -->     Srpen 2002<--
manpagestest/target/intro.1.po:171: (po4a::po)
Strange line: -->Ne Po Út St Čt Pá So<--
manpagestest/target/intro.1.po:172: (po4a::po)
Strange line: -->             1  2  3<--
manpagestest/target/intro.1.po:173: (po4a::po)
Strange line: --> 4  5  6  7  8  9 10<--
manpagestest/target/intro.1.po:174: (po4a::po)
Strange line: -->11 12 13 14 15 16 17<--
manpagestest/target/intro.1.po:175: (po4a::po)
Strange line: -->18 19 20 21 22 23 24<--
manpagestest/target/intro.1.po:176: (po4a::po)
Strange line: -->25 26 27 28 29 30 31<--
manpagestest/target/intro.1.po:236: (po4a::po)
Strange line: --><--
manpagestest/target/intro.1.po:309: (po4a::po)
Strange line: --><--
...

as you can see, po4a complains about some extra blank lines in the file and also does not like at all the translated part. This is because po4a expects the message to have opening and closing doublequotes, and to have them on each line (a bug in OmegaT?). I fix that manually:

$ cp manpagestest/target/intro.1.po manpagestest/target/intro.1.fixed.po
$ vi manpagestest/target/intro.1.fixed.po
$ cat manpagestest/target/intro.1.fixed.po
...
# type: Plain text
#: en/intro.1:106
#, no-wrap
msgid ""
"B<knuth login: >I<aeb>\n"
"B<Password: >I<********>\n"
"B<% >I<date>\n"
"Tue Aug  6 23:50:44 CEST 2002\n"
"B<% >I<cal>\n"
"     August 2002\n"
"Su Mo Tu We Th Fr Sa\n"
"             1  2  3\n"
" 4  5  6  7  8  9 10\n"
"11 12 13 14 15 16 17\n"
"18 19 20 21 22 23 24\n"
"25 26 27 28 29 30 31\n"
msgstr "B<knuth login: >I<aeb>"
"B<Password: >I<********>"
"B<% >I<date>"
"Út Srp  6 23:50:44 CEST 2002"
"B<% >I<cal>"
"     Srpen 2002"
"Ne Po Út St Čt Pá So"
"             1  2  3"
" 4  5  6  7  8  9 10"
"11 12 13 14 15 16 17"
"18 19 20 21 22 23 24"
"25 26 27 28 29 30 31"

...

and retry:

$ po4a-translate -f man -m en/intro.1 -p manpagestest/target/intro.1.fixed.po -l intro.cs.1

The important part of the created intro.cs.1 looks like this:

...
\fBknuth login: \fP\fIaeb\fP\fBPassword: \fP\fI********\fP\fB% \fP\fIdate\fPÚt Srp  6 23:50:44 CEST 2002\fB% \fP\fIcal\fP     Srpen 2002Ne Po Út St Čt Pá So             1  2  3 4  5  6  7  8  9 1011 12 13 14 15 16 1718 19 20 21 22 23 2425 26 27 28 29 30 31
...

and this is how it looks when viewed with man ./intro.cs.1, i.e. badly:

All the mentioned files are attached.

       

Obrázky

Komentáře

[hajma]

Příloha:

• bugreport.tar.bz2 (18563 bytů)

All the mentioned files are in the attachment.

[sssssssss]

A ty si to nemůžeš uložit jinam než na abíčko?

[Limoto]

Třeba do bugzilly ábíčka

[hajma]

nó, ábíčko je pro mě nejpohodlnější

samozřejmě jsem měl radši napsat něco o ruské invazi, ale být offtopic mi občas vyhovuje

[pavlix]

No comment, chci být slušný :).

[kotyz]

Nejde udelat aby tejden zacinal pondelim?

[hajma]

všechno jde. Manuálová stránka nicméně odráží skutečný stav věcí. Zkus nahlásit bug v util-linux-ng, kam příkaz cal patří.

[disorder]

cal -m

[hajma]

ale bylo by fajn, kdyby si to cal našel podle locale sám

[disorder]

neviem o ziadnom LC_DATE

[David Watzke]

AFAIK k tomu slouží LC_TIME.

[disorder]

cas sa formatuje oddelene, datum sa formatuje oddelene a toto je este nieco uplne ine

co by si nastavil, keby si chcel 24h cas a americke datumy?

[disorder]

a vobec, do toho by som to ani ja nezahrnul, s formatom datumu to tiez nema nic

[petr_p]

Jde, dokonce to ve zdrojích je zakomentované. Jenže vyšší místa rozhodla, že by chování podle locale bylo v rozporu s manuálem a, co je horší, údajně i s POSIXem.

Kdyby to někoho zajímalo, můžu sem hodit patch. Někde se mi válí na disku.

[petr_p]

Výborně. Konečně rozumný způsob, jak překládat manuál.

Nicméně spárování originálu a překladu umí i po4a. Je přínosné se zlobit s OmegouT?