Differential D16821
Add an optional pht() callback for collecting string frequencies Authored by epriestley on Nov 8 2016, 2:49 PM. Tags None Referenced Files
Subscribers None
Details
Ref T5267. For translations, it's useful to have frequency data so the most frequently used strings can be prioritized for translation -- common UI strings are much more important to translate than big blocks of setup/configuration text or obscure error messages. Add an optional callback so we can collect which strings are actually translated at runtime and build a dataset for prioritizing strings for translation. Applied this simple hack to phabricator/: diff --git a/src/infrastructure/env/PhabricatorEnv.php b/src/infrastructure/env/PhabricatorEnv.php index 31fd116..9e08a21 100644 --- a/src/infrastructure/env/PhabricatorEnv.php +++ b/src/infrastructure/env/PhabricatorEnv.php @@ -59,6 +59,8 @@ final class PhabricatorEnv extends Phobject { private static $readOnly; private static $readOnlyReason; + public static $frequency = array(); + const READONLY_CONFIG = 'config'; const READONLY_UNREACHABLE = 'unreachable'; const READONLY_SEVERED = 'severed'; @@ -166,7 +168,8 @@ final class PhabricatorEnv extends Phobject { PhutilTranslator::getInstance() ->setLocale($locale) - ->setTranslations($override + $translations); + ->setTranslations($override + $translations) + ->setWillTranslateCallback('PhabricatorEnv::willTranslate'); self::$localeCode = $locale_code; } catch (Exception $ex) { @@ -174,6 +177,13 @@ final class PhabricatorEnv extends Phobject { } } + public static function willTranslate($text) { + if (empty(self::$frequency[$text])) { + self::$frequency[$text] = 0; + } + self::$frequency[$text]++; + } + private static function buildConfigurationSourceStack($config_optional) { self::dropConfigCache(); diff --git a/webroot/index.php b/webroot/index.php index 59e5b71..d57dd35 100644 --- a/webroot/index.php +++ b/webroot/index.php @@ -15,6 +15,14 @@ try { try { PhabricatorStartup::beginStartupPhase('run'); AphrontApplicationConfiguration::runHTTPRequest($sink); + + $freq = PhabricatorEnv::$frequency; + asort($freq); + Filesystem::writeFile( + '/tmp/pht_frequency.json', + id(new PhutilJSON()) + ->encodeFormatted($freq)); + } catch (Exception $ex) { try { $response = new AphrontUnhandledExceptionResponse(); Got a datafile like this: { "Piece of Eight": 1, "Haypence": 1, "Disable DarkConsole": 1, "Yellow Medal": 1, "Doubloon": 1, "Mountain of Wealth": 1, "Baby Tequila": 1, "Evil Spooky Haunted Tree": 1, "Pterodactyl": 1, ...
Diff Detail
Event TimelineComment Actions We sampled this at FB like multimeter, to see which showed up the most in real usage. Comment Actions Yeah, I'm planning to do something similar here, although I'll probably just run it at 100% on this install for a week or something to start with. I don't think we'll get much better/different data by doing something super complicated with the Phacility cluster, although we could do that eventually. Comment Actions That is, I'll sample in phabricator/ on a per-request basis like this: if (mt_rand() % 12345 == 0) {
$translator-> setWillTranslateCallback(...);
}If we sample in libphutil/, we'd have a bunch of overhead on every call to pht(). |