jenkins-bot has submitted this change and it was merged.
Change subject: Add support for TMX level 1 (plaintext)
......................................................................
Add support for TMX level 1 (plaintext)
I think this is enough and level 2 is not worth it.
JSON format is superior as it contains MT output as well.
Bug: T122042
Change-Id: Ia9f75cf45fc89e3b25054bc367fa47dec60352b3
---
M scripts/dump-corpora.php
1 file changed, 111 insertions(+), 21 deletions(-)
Approvals:
Santhosh: Looks good to me, approved
Nikerabbit: Looks good to me, but someone else must approve
jenkins-bot: Verified
diff --git a/scripts/dump-corpora.php b/scripts/dump-corpora.php
index 350b199..2399c84 100644
--- a/scripts/dump-corpora.php
+++ b/scripts/dump-corpora.php
@@ -40,7 +40,7 @@
$this->addOption(
'format',
- '(optional) Dump format. Defaults to JSON.',
+ '(optional) Dump format. Available formats json
(default) and tmx.',
false, /*required*/
true /*has arg*/
);
@@ -69,6 +69,8 @@
$split = $this->getOption( 'split-at', false );
$type = $plain ? 'text' : 'html';
+ $formatSpec = array( $format, $type );
+
$limit = 999999999;
$offset = 0;
$translations = Translation::getAllPublishedTranslations(
@@ -83,16 +85,30 @@
$lookup = new CorporaLookup( $db );
foreach ( $translations as &$translation ) {
$translation['corpora'] = $lookup->getByTranslationId(
$translation['translationId'] );
- if ( $plain ) {
- foreach ( $translation['corpora'] as $id =>
$unit ) {
- foreach ( $unit as $field => $value ) {
- if ( !isset( $value['content']
) ) {
- continue;
- }
-
$translation['corpora'][$id][$field]['content'] =
-
Sanitizer::stripAllTags( $value['content'] );
+ // Some general cleanup
+ foreach ( $translation['corpora'] as $id => $unit ) {
+ if ( !isset( $unit['user'] ) ) {
+ unset( $translation['corpora'][$id] );
+ continue;
+ }
+
+ unset(
$translation['corpora'][$id]['source']['engine'] );
+ unset(
$translation['corpora'][$id]['user']['engine'] );
+ }
+
+ if ( !$plain ) {
+ continue;
+ }
+
+ foreach ( $translation['corpora'] as $id => $unit ) {
+ foreach ( $unit as $field => $value ) {
+ if ( !isset( $value['content'] ) ) {
+ continue;
}
+
+
$translation['corpora'][$id][$field]['content'] =
+ Sanitizer::stripAllTags(
$value['content'] );
}
}
}
@@ -101,7 +117,7 @@
$source = $sourceLanguage ?: '_';
$target = $targetLanguage ?: '_';
$filename =
"cx-corpora.{$source}2{$target}.$type.$format";
- $this->export( $format, $filename, $translations );
+ $this->export( $formatSpec, $filename, $translations,
$source );
return;
}
@@ -114,7 +130,7 @@
}
$filename =
"cx-corpora.{$sourceLanguage}2{$targetLanguage}.$type.$format";
- $this->export( $format, $filename, $targets );
+ $this->export( $formatSpec, $filename,
$targets, $sourceLanguage );
unset(
$sorted[$targetLanguage][$sourceLanguage] );
}
@@ -132,14 +148,14 @@
}
$filename =
"cx-corpora._2{$targetLanguage}.$type.$format";
- $this->export( $format, $filename, $targets );
+ $this->export( $formatSpec, $filename, $targets, '_' );
unset( $sorted[$targetLanguage] );
}
if ( count( $sorted ) ) {
$targets = call_user_func_array( 'array_merge', $sorted
);
$filename = "cx-corpora._2_.$type.$format";
- $this->export( $format, $filename, $targets );
+ $this->export( $formatSpec, $filename, $targets, '_' );
}
}
@@ -163,27 +179,33 @@
return $sorted;
}
- public function export( $format, $filename, array $targets ) {
- if ( $format !== 'json' ) {
+ public function export( $formatSpec, $filename, array $targets,
$sourceLanguage ) {
+ $data = null;
+
+ list( $format, $type ) = $formatSpec;
+
+ if ( $format === 'json' ) {
+ $data = $this->formatJSON( $targets );
+ } elseif ( $format === 'tmx' ) {
+ $data = $this->formatTMX( $targets, $sourceLanguage,
$type );
+ } else {
$this->error( "Unknown output format\n", 1 );
}
- $data = $this->formatJSON( $targets );
if ( $data ) {
file_put_contents( $filename, $data );
$this->output( "$filename\n" );
}
}
+ /**
+ * @param array $targets
+ * @return string|null
+ */
public function formatJSON( array $targets ) {
$output = array();
foreach ( $targets as $translation ) {
foreach ( $translation['corpora'] as $id => $unit ) {
- if ( !isset( $unit['user'] ) ) {
- continue;
- }
-
- unset( $unit['source']['engine'],
$unit['user']['engine'] );
unset( $unit['source']['timestamp'],
$unit['user']['timestamp'], $unit['mt']['timestamp'] );
$globalId =
"{$translation['translationId']}/$id";
@@ -204,6 +226,74 @@
return null;
}
}
+
+ /**
+ * @param array $targets
+ * @param string $sourceLanguage Language code.
+ * @param string $type Either html or plain.
+ * @return string|null
+ */
+ public function formatTMX( array $targets, $sourceLanguage, $type ) {
+ if ( $type === 'html' ) {
+ $this->error( "TMX output format is only supported with
plaintext\n", 1 );
+ }
+
+ $xml = new DOMDocument( '1.0', 'UTF-8' );
+ $spawn = function ( $tag, array $attributes = array() ) use (
$xml ) {
+ $element = $xml->createElement( $tag );
+ foreach ( $attributes as $key => $value ) {
+ $element->setAttribute( $key, $value );
+ }
+
+ return $element;
+ };
+
+ $tmx = $spawn( 'tmx', array( 'version' => '1.4' ) );
+ $xml->appendChild( $tmx );
+
+ $header = $spawn( 'header', array(
+ 'creationtool' => 'dump-corpora.php / DOMDocument',
+ 'creationtoolversion' => '1.0.0',
+ // Could be paragraph, but not guaranteed...
+ 'segtype' => 'block',
+ 'o-tmf' => 'sql',
+ 'adminlang' => 'en',
+ 'sourcelang' => $sourceLanguage === '_' ? '*all*' :
$sourceLanguage,
+ 'datatype' => 'plaintext',
+ ) );
+ $tmx->appendChild( $header );
+
+ $body = $spawn( 'body' );
+ $tmx->appendChild( $body );
+
+ foreach ( $targets as $translation ) {
+ foreach ( $translation['corpora'] as $id => $units ) {
+ $tu = $spawn( 'tu', array( 'srclang' =>
$translation['sourceLanguage'] ) );
+ $body->appendChild( $tu );
+ foreach ( $units as $origin => $unit ) {
+ if ( $unit['content'] === null ) {
+ continue;
+ }
+ if ( $origin === 'source' ) {
+ $tuv = $spawn( 'tuv', array(
'xml:lang' => $translation['sourceLanguage'] ) );
+ } else {
+ $tuv = $spawn( 'tuv', array(
'xml:lang' => $translation['targetLanguage'] ) );
+ }
+ $tu->appendChild( $tuv );
+ $tuvProp = $spawn( 'prop', array(
'type' => 'origin' ) );
+ $tuvProp->appendChild(
$xml->createTextNode( $origin ) );
+ $tuv->appendChild( $tuvProp );
+ $tuvSeg = $spawn( 'seg' );
+ $tuv->appendChild( $tuvSeg );
+ $tuvContent = $xml->createTextNode(
$unit['content'] );
+ $tuvSeg->appendChild( $tuvContent );
+ }
+ }
+ }
+ // Format the output with indentation for readability
+ $xml->formatOutput = true;
+ return $xml->saveXML();
+ }
}
$maintClass = 'CXCorporaDump';
--
To view, visit https://gerrit.wikimedia.org/r/272461
To unsubscribe, visit https://gerrit.wikimedia.org/r/settings
Gerrit-MessageType: merged
Gerrit-Change-Id: Ia9f75cf45fc89e3b25054bc367fa47dec60352b3
Gerrit-PatchSet: 5
Gerrit-Project: mediawiki/extensions/ContentTranslation
Gerrit-Branch: master
Gerrit-Owner: Nikerabbit <[email protected]>
Gerrit-Reviewer: KartikMistry <[email protected]>
Gerrit-Reviewer: Nikerabbit <[email protected]>
Gerrit-Reviewer: Santhosh <[email protected]>
Gerrit-Reviewer: jenkins-bot <>
_______________________________________________
MediaWiki-commits mailing list
[email protected]
https://lists.wikimedia.org/mailman/listinfo/mediawiki-commits
