Revision history
1.4 July 30, 2019
- Support `--lang` parameter in `serge pull-ts` and `serge push-ts`
(gives an ability to specify one or more languages which should be
synchronized with the translation service). Previously this parameter
worked only for `serge sync` and `serge localize` and only affected
the localization step.
- Properly handle `leave_untranslated_blank` job option in .properties,
.ts, .rrc and .strings parsers (previously this option wouldn't affect
untranslated strings in the generated localized files leaving source
string values instead).
- POT parser can now generate compiled .mo files in addition to producing
localized .po files.
- Make Serge installable via `./Build install`
- Git plugin now always uses `reset --hard` instead of `rebase`
when doing updates so that it can recover from history rewrites
on a remote server.
- Better error reporting when spawning child processes.
- `output_default_lang_file` config setting now works correctly with
the `limit_languages` plugin. Additionally, the list of target
languages is now de-duplicated and sorted before job runs.
- Implemented cache eviction optimizations to make sure the process
uses less memory on large databases.
- Implemented four different caching strategies ('db', 'namespace',
'file', and 'string') that can be set with SERGE_DB_CACHING_STRATEGY
environment variable. The more granular the level, the less memory
Serge uses, at the potential expense of longer execution time.
The default is 'db', which is backward-compatible with the old
behavior where the entire database would be cached in memory.
- `add_dev_comment` callback is now deprecated in favor of `add_hint`
callback which provides more parameters than the old one.
- New `keys_language` plugin that allows one to generate globally
unique keys in place of translation. An ability to uniquely address
any string opens possibilities for some advanced techniques
like dynamic localization and integration with external tools.
- parse_ts parser improved to support proper string literal
escaping/unescaping, which now supports sequences like
\NNN, \xNN, \uNNNN, \u{NNNNN}. \n and \t sequences are properly
converted to raw line breaks and tabulation symbols, respectively.
- parse_master can now be configured to use custom opening/closing
markers and delimiters.
- Added feature branch config generator tool (see
`bin/tools/feature-branch-config-generator` folder for the tool itself,
examples and documentation). This tool scans your Git repository,
determines the list of actively maintained branches that match the
provided name masks, and generates a Serge config to localize all
these branches based on the provided config template.
- `if` plugin can now capture content using regular expressions
and keep it around for child plugin classes, as well as use them
in test statements (`has_capture`, `has_no_capture`, `has_all_captures`).
`append_hint_message` plugin can now replace `%CAPTURE:...%` macros
with the values of such captures. This is useful, for example, to extract
a specific string from the header of the source file and expose it
as a hint (comment) in all translation units.
- Net::SMTP is now used instead of deprecated Net::SMTP::SSL.
- An ability to use XLIFF 1.2 as a translation interchange format
added (previously the only supported interchange format was .po).
Using XLIFF for translation might come in handy when working with
certain CAT tools or online services.
- Added `rewrite_relative_ts_file_path` and `rewrite_absolute_ts_file_path`
phases that allow one to manipulate TS file names with plugins.
- Added `after_serialize_ts_file` and `before_deserialize_ts_file`
phases that allow one to manipulate the translation interchange
file contents.
- Added `apply_xslt` plugin, which allows one to apply XSLT transformations
to source or target XML resource files, as well as XLIFF
translation interchange files.
- Split `pootle` and `zing` translation plugins for better clarity.
- Git sync plugin now supports 'fetch_params',
`commit_params` and `push_params` configuration parameters
in addition to a previously existed `clone_params`. This
allows one to pass extra parameters to `git` during different
stages of its work. This is useful for e.g. shallow cloning
and fetching, or for bypassing commit and push hooks.
- Improved `transform` plugin logic to make trailing punctuation
detection more smart when it is combined with wrapper tags.
It also deals better with whitespace, reducing the list of
candidates to guess transformations for.
- `transform` plugin now always returns guessed translations as fuzzy;
its previous configuration parameters (`as_fuzzy_default`,
`as_fuzzy` and `as_not_fuzzy`) are now deprecated. You will see
a corresponding notice in the command line output if you still
have these parameters in your configuration file.
- Fixed a bug where `reuse_orphaned` job parameter would not be passed down
to the code that expected it, which resulted in not reusing orphaned
translations in the `transform` plugin.
- Refactored `feature_branch` plugin to more accurately reuse
master branch translations in dependent branches.
This required to selectively disable optimizations when localizing
master files, so handling localization with `feature_branch`
might become a bit slower.
- `parse_keyvalue` plugin now exports keys explicitly
(in addition to extracting them as translation hints).
- Added support for `use_keys_as_context` job parameter (#89).
- Added new `show` command to display expanded .serge configurations (#40).
- Miscellaneous smaller fixes and improvements.
1.3 February 1, 2018
- Fix bug when extra item comment would be appended only if the
original comment is not blank
- In interactive console mode, `serge help` tries to use a pager,
e.g. `less` or `more` (#39)
- In POT parser, bypass non-standard comments that other tools
may produce (#41)
- Add summary at the end of the sync command: how many
jobs and configs were processed, how many configs and jobs
were skipped due to errors, and how many jobs ended abnormally.
This will help easily identify issues both from the command line
and when analyzing logs
- JSON parser now supports parsing nested HTML-formatted values
(like XML parser already does)
- PHP/XHTML parser now treats bare HTML strings (with no wrapper tag)
as translatable by default, except when they contain PHP tags.
This is useful when parsing nested HTML snippets inside XML or JSON
- Parsing errors in nested HTML values inside JSON or XML are now
exposed to the parent parser for proper error reporting
- `test-parser` command now accepts `--as-object` flag to produce
a more verbose and self-descriptive output (an array of objects)
- JSON and XML parsers can now have a nested HTML parser configured
via `html_parser` parameter (one can specify both parser name
and its parameters). If this parameter is not provided, parsers
fall back to a previous behavior by using `parse_php_xhtml`
with the default parameters as a nested parser
- New `rewrite_source` callback to rewrite source strings on the fly
before they are passed through the translation pipeline
- If the parser does not return key names for `serge import`, it is now
possible to use `--disambiguate-keys` to auto-generate unique key names
and do the import
- Qt Linguist TS Parser has been rewritten as a proper validating
XML parser. It now also extracts context name, comments,
and supports import mode
- Translations containing just one '0' symbol are now properly saved to
database.
- Rendering of units with notices/warnings/errors is now fixed
in `serge import` HTML reports.
- parse_js plugin now correctly handles lines with end-of-line comments
containing quotes.
- Git sync plugin now supports `clone_params` configuration parameter
that allows one to pass extra parameters on initial project checkout
(useful for e.g. shallow cloning).
- .PO serializer now normalizes line endings prior to parsing the file.
This allows one to import .PO files with Windows-style line endings.
1.2 October 29, 2016
- Make TS File serialization pluggable (see #11)
- .PO file support is now implemented as a serialization plugin
- .CSV serialization plugin added
- Fix feature_branch plugin to properly compare strings against the master job,
not the entire database
- Fix parse_pot plugin bug in import mode (language would not be passed back
in a callback in import mode, causing a DB constraint error)
- Print nice plural separators and show line breaks in strings in import reports
- JSON parser now accepts source files with relaxed syntax (trailing commas and
comments) (#13)
- Fix bug in translation variant counting that would prevent reusing strings
effectively (incorrectly treating more translations as "uncertain" ones)
- When MySQL database is initialized, utf8_bin collation is now enforced
- parse_js plugin can now extract comments at the end of each line
- Add `serge import` support for .plist parser (#16)
- Add `serge import` support for parse_json and parse_json_keyvalue parsers
- Improve `transform` plugin logic
- Add `yaml_kind` to the YAML parser to support Rails-specific file format
- Add `leave_untranslated_blank` job option to allow cleaning up untranslated
entries during post-processing; this in turn allows for custom fallback
mechanisms and also reduces resource file sizes
- In Git plugin, old branches are automatically pruned to avoid conflicts
on `git fetch`
- JSON and YAML parsers now sort keys alphabetically at parsing time.
This is to guarantee stable output between multiple runs and between
different Perl versions (#29)
- 'before_update_database_from_ts_file' callback has been deprecated in favor of
'before_update_database_from_ts_files'
- Fix unwanted escaping of quotes/apostrophes inside tag markup embedded in a string
(e.g. <xliff:g id="...">)
- Added support for tests that report job validation errors (and compare them
with reference error messages)
- Added better error reporting for incorrectly set up jobs, e.g. the ones
that have empty `db_source` or `db_namespace` parameters
- `db_namespace` parameter can now contain %ENV:...% macros
- When job fails, we now move on to the next job in the config
(previously the entire config would be skipped)
- Added version information to the CLI via the `--version` flag (#38)
- Added support for all escaped characters as defined in .po format (#44)
1.1 April 20, 2016
- On OS X, expand `~` to the actual `$HOME` path in SQLite DB connection string (#1)
- Accept multiple `message` entries in append_hint_message plugin (#2)
- Require JSON::XS (#3)
- Make serge and tools find their libs when they are symlinked (#4)
- Fixed: Uppercase letters in language names are unsupported (#5)
- Accept multiple `command` entries in run_command plugin (#6)
- parse_android plugin now supports `serge import` mode
(generates string keys for alignment) (#7)
- parse_pot plugin now supports `serge import` mode
(generates string keys for alignment) (#8)
- Fix performance when `reuse_uncertain` mode is off (#9)
- parse_xliff plugin now supports `serge import` mode
- Parsers are now aware when they are used in import mode, and can adjust their
behavior accordingly; one can now run `serge test-parser --import-mode <...>`
to test parsers in import mode.
- Added the ability to specify any arbitrary class name and plugin data for
`test-parser` command, which allows one to use this command for any
custom plugins
- Added support for %OUTFILE% and %OUTPATH% macros in `run_command` plugin;
other macros like %FILE% now work consistently with the rest of the config code
- Fixed some tests due to broken source files
- Fixed a regression with job's `active` parameter not defaulting to `YES`
- Added `feature_branch` plugin that allows one to add strings from multiple sources
(branches) and exclude duplicate strings.
- Added `log_translation` callback that allows one to log/cache translations
- String context and key is now passed in `can_extract` callback
- String key is now passed in `get_translation` and `get_translation_pre` callbacks
- `test_language` plugin now handles URLs better (doesn't try to
transliterate them), it allows string expansion (disabled by default),
and transliteration can now be optionally turned off (it is enabled by default)
- `test_language` plugin now won't generate TS files if its `save_translations`
option is turned off
- `completeness` plugin has got a new `bypass_languages` option the allows to
skip its logic for certain languages (this is primarily useful for
pseudo-localization, see `test_language` plugin)
- Caching/lookup algorithm for translations was rewritten, giving a significant
performance boost when processing new strings, at the expense of loading
more data into memory (which also can take some time depending on a size of the
database)
- Fixed USN-based optimization when generating TS files
- Fixed TS file content hash-based optimization caused by inconsistent UTF-8 handling
at file save/load time
- Fixed text unescaping in XML parser (it is now CDATA-aware and is also applied
for embedded HTML nodes)
- `limit_languages` plugin now properly matches languages followed by `-->`
(when they are embedded as HTML comments) and correctly works with
negative rules (`file_doesnt_match`, `content_doesnt_match`).
It also now has its default matching rules properly overridable.
- `serge clean-ts` now properly initializes job defaults so its behavior
now matches that of `serge localize` and `serge import`, and also
works properly with symlinked files/folders when scanning the directory
tree.
- Fixed %LOCALE:LC% and %LOCALE:UC% macros that would be affected by
iOS-specific language remappings (e.g. pt and pt-br).
- Added new `serge show <config>` command to expand the @inherit
rules in the configuration file and render the nicely formatted
final config structure.
- Fixed typos in embedded documentation
1.0 October 15, 2015
* Code published on GitHub *