195640e3a20adea634b4df4ccf8c93f411184c438joi@chromium.org#!/usr/bin/env python 295640e3a20adea634b4df4ccf8c93f411184c438joi@chromium.org# Copyright (c) 2012 The Chromium Authors. All rights reserved. 301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Use of this source code is governed by a BSD-style license that can be 401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# found in the LICENSE file. 501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org'''A gatherer for the TotalRecall brand of HTML templates with replaceable 701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgportions. We wanted to reuse extern.tclib.api.handlers.html.TCHTMLParser 801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgbut this proved impossible due to the fact that the TotalRecall HTML templates 901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgare in general quite far from parseable HTML and the TCHTMLParser derives 1001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom HTMLParser.HTMLParser which requires relatively well-formed HTML. Some 1101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgexamples of "HTML" from the TotalRecall HTML templates that wouldn't be 1201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgparseable include things like: 1301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 1401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <a [PARAMS]>blabla</a> (not parseable because attributes are invalid) 1501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 1601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <table><tr><td>[LOTSOFSTUFF]</tr></table> (not parseable because closing 1701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org </td> is in the HTML [LOTSOFSTUFF] 1801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org is replaced by) 1901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgThe other problem with using general parsers (such as TCHTMLParser) is that 2101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgwe want to make sure we output the TotalRecall template with as little changes 2201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgas possible in terms of whitespace characters, layout etc. With any parser 2301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgthat generates a parse tree, and generates output by dumping the parse tree, 2401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgwe would always have little inconsistencies which could cause bugs (the 2501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgTotalRecall template stuff is quite brittle and can break if e.g. a tab 2601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgcharacter is replaced with spaces). 2701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgThe solution, which may be applicable to some other HTML-like template 2901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orglanguages floating around Google, is to create a parser with a simple state 3001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgmachine that keeps track of what kind of tag it's inside, and whether it's in 3101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orga translateable section or not. Translateable sections are: 3201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 3301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orga) text (including [BINGO] replaceables) inside of tags that 3401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org can contain translateable text (which is all tags except 3501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for a few) 3601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 3701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgb) text inside of an 'alt' attribute in an <image> element, or 3801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org the 'value' attribute of a <submit>, <button> or <text> 3901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org element. 4001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 4101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgThe parser does not build up a parse tree but rather a "skeleton" which 4201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgis a list of nontranslateable strings intermingled with grit.clique.MessageClique 4301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgobjects. This simplifies the parser considerably compared to a regular HTML 4401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgparser. To output a translated document, each item in the skeleton is 4501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgprinted out, with the relevant Translation from each MessageCliques being used 4601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfor the requested language. 4701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 4801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgThis implementation borrows some code, constants and ideas from 4901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgextern.tclib.api.handlers.html.TCHTMLParser. 5001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org''' 5101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 5201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 5301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgimport re 5401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgimport types 5501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 5601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom grit import clique 5701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom grit import exception 5801fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.orgfrom grit import lazy_re 5901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom grit import util 6001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom grit import tclib 6101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 6201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgfrom grit.gather import interface 6301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 6401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 6501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# HTML tags which break (separate) chunks. 6601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_BLOCK_TAGS = ['script', 'p', 'h1', 'h2', 'h3', 'h4', 'h5', 'h6', 'hr', 'br', 6701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'body', 'style', 'head', 'title', 'table', 'tr', 'td', 'th', 6801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'ul', 'ol', 'dl', 'nl', 'li', 'div', 'object', 'center', 6901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'html', 'link', 'form', 'select', 'textarea', 7001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'button', 'option', 'map', 'area', 'blockquote', 'pre', 7101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'meta', 'xmp', 'noscript', 'label', 'tbody', 'thead', 728a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 'script', 'style', 'pre', 'iframe', 'img', 'input', 'nowrap', 738a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 'fieldset', 'legend'] 7401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 7501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# HTML tags which may appear within a chunk. 7601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_INLINE_TAGS = ['b', 'i', 'u', 'tt', 'code', 'font', 'a', 'span', 'small', 7701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'key', 'nobr', 'url', 'em', 's', 'sup', 'strike', 7801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'strong'] 7901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 8001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# HTML tags within which linebreaks are significant. 8101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_PREFORMATTED_TAGS = ['textarea', 'xmp', 'pre'] 8201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 8301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# An array mapping some of the inline HTML tags to more meaningful 8401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# names for those tags. This will be used when generating placeholders 8501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# representing these tags. 8601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_HTML_PLACEHOLDER_NAMES = { 'a' : 'link', 'br' : 'break', 'b' : 'bold', 8701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'i' : 'italic', 'li' : 'item', 'ol' : 'ordered_list', 'p' : 'paragraph', 8801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'ul' : 'unordered_list', 'img' : 'image', 'em' : 'emphasis' } 8901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 9001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# We append each of these characters in sequence to distinguish between 9101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# different placeholders with basically the same name (e.g. BOLD1, BOLD2). 9201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Keep in mind that a placeholder name must not be a substring of any other 9301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# placeholder name in the same message, so we can't simply count (BOLD_1 9401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# would be a substring of BOLD_10). 9501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_SUFFIXES = '123456789ABCDEFGHIJKLMNOPQRSTUVWXYZ' 9601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 9701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches whitespace in an HTML document. Also matches HTML comments, which are 9801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# treated as whitespace. 9901fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_WHITESPACE = lazy_re.compile(r'(\s| |\\n|\\r|<!--\s*desc\s*=.*?-->)+', 10001fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org re.DOTALL) 10101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 1028a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org# Matches whitespace sequences which can be folded into a single whitespace 1038a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org# character. This matches single characters so that non-spaces are replaced 1048a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org# with spaces. 1058a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org_FOLD_WHITESPACE = lazy_re.compile(r'\s+') 1068a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 10701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Finds a non-whitespace character 10801fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_NON_WHITESPACE = lazy_re.compile(r'\S') 10901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 11001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches two or more in a row (a single   is not changed into 11101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# placeholders because different languages require different numbers of spaces 11201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# and placeholders must match exactly; more than one is probably a "special" 11301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# whitespace sequence and should be turned into a placeholder). 11401fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_NBSP = lazy_re.compile(r' ( )+') 11501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 11601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches nontranslateable chunks of the document 11701fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_NONTRANSLATEABLES = lazy_re.compile(r''' 11801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*script.+?<\s*/\s*script\s*> 11901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 12001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*style.+?<\s*/\s*style\s*> 12101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 12201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <!--.+?--> 12301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 12401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\?IMPORT\s.+?> # import tag 12501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 12601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*[a-zA-Z_]+:.+?> # custom tag (open) 12701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 12801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*/\s*[a-zA-Z_]+:.+?> # custom tag (close) 12901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 13001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <!\s*[A-Z]+\s*([^>]+|"[^"]+"|'[^']+')*?> 13101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''', re.MULTILINE | re.DOTALL | re.VERBOSE | re.IGNORECASE) 13201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 13301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches a tag and its attributes 13401fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_ELEMENT = lazy_re.compile(r''' 13501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Optional closing /, element name 13601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*(?P<closing>/)?\s*(?P<element>[a-zA-Z0-9]+)\s* 13701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Attributes and/or replaceables inside the tag, if any 13801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org (?P<atts>( 13901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org \s*([a-zA-Z_][-:.a-zA-Z_0-9]*) # Attribute name 14001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org (\s*=\s*(\'[^\']*\'|"[^"]*"|[-a-zA-Z0-9./,:;+*%?!&$\(\)_#=~\'"@]*))? 14101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 14201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org \s*\[(\$?\~)?([A-Z0-9-_]+?)(\~\$?)?\] 14301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org )*) 14401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org \s*(?P<empty>/)?\s*> # Optional empty-tag closing /, and tag close 14501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''', 14601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org re.MULTILINE | re.DOTALL | re.VERBOSE) 14701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 14801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches elements that may have translateable attributes. The value of these 14901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# special attributes is given by group 'value1' or 'value2'. Note that this 15001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# regexp demands that the attribute value be quoted; this is necessary because 15101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# the non-tree-building nature of the parser means we don't know when we're 15201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# writing out attributes, so we wouldn't know to escape spaces. 15301fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_SPECIAL_ELEMENT = lazy_re.compile(r''' 15401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <\s*( 15501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org input[^>]+?value\s*=\s*(\'(?P<value3>[^\']*)\'|"(?P<value4>[^"]*)") 15601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org [^>]+type\s*=\s*"?'?(button|reset|text|submit)'?"? 15701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 15801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ( 15901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org table[^>]+?title\s*= 16001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 16101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org img[^>]+?alt\s*= 16201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 16301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org input[^>]+?type\s*=\s*"?'?(button|reset|text|submit)'?"?[^>]+?value\s*= 16401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ) 16501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org \s*(\'(?P<value1>[^\']*)\'|"(?P<value2>[^"]*)") 16601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org )[^>]*?> 16701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''', re.MULTILINE | re.DOTALL | re.VERBOSE | re.IGNORECASE) 16801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 16901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches stuff that is translateable if it occurs in the right context 17001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# (between tags). This includes all characters and character entities. 17101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Note that this also matches which needs to be handled as whitespace 17201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# before this regexp is applied. 17301fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_CHARACTERS = lazy_re.compile(r''' 17401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ( 17501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org \w 17601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 17701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org [\!\@\#\$\%\^\*\(\)\-\=\_\+\[\]\{\}\\\|\;\:\'\"\,\.\/\?\`\~] 17801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org | 17901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org &(\#[0-9]+|\#x[0-9a-fA-F]+|[A-Za-z0-9]+); 18001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org )+ 18101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''', re.MULTILINE | re.DOTALL | re.VERBOSE) 18201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 18301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches Total Recall's "replaceable" tags, which are just any text 18401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# in capitals enclosed by delimiters like [] or [~~] or [$~~$] (e.g. [HELLO], 18501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# [~HELLO~] and [$~HELLO~$]). 18601fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_REPLACEABLE = lazy_re.compile(r'\[(\$?\~)?(?P<name>[A-Z0-9-_]+?)(\~\$?)?\]', 18701fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org re.MULTILINE) 18801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 18901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 19001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches the silly [!]-prefixed "header" that is used in some TotalRecall 19101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# templates. 19201fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_SILLY_HEADER = lazy_re.compile(r'\[!\]\ntitle\t(?P<title>[^\n]+?)\n.+?\n\n', 19301fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org re.MULTILINE | re.DOTALL) 19401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 19501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 19601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org# Matches a comment that provides a description for the message it occurs in. 19701fadb72b6e94e6511eaffd1874a8cc095f098a7joi@chromium.org_DESCRIPTION_COMMENT = lazy_re.compile( 19801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org r'<!--\s*desc\s*=\s*(?P<description>.+?)\s*-->', re.DOTALL) 19901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2008a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org# Matches a comment which is used to break apart multiple messages. 2018a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org_MESSAGE_BREAK_COMMENT = lazy_re.compile(r'<!--\s*message-break\s*-->', 2028a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org re.DOTALL) 2038a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 204807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org# Matches a comment which is used to prevent block tags from splitting a message 205807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org_MESSAGE_NO_BREAK_COMMENT = re.compile(r'<!--\s*message-no-break\s*-->', 206807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org re.DOTALL) 207807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org 20801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 20901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org_DEBUG = 0 21001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgdef _DebugPrint(text): 21101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if _DEBUG: 21201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org print text.encode('utf-8') 21301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 21401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 21501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgclass HtmlChunks(object): 21601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''A parser that knows how to break an HTML-like document into a list of 21701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org chunks, where each chunk is either translateable or non-translateable. 21801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org The chunks are unmodified sections of the original document, so concatenating 21901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org the text of all chunks would result in the original document.''' 22001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 22101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def InTranslateable(self): 22201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return self.last_translateable != -1 22301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 22401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def Rest(self): 22501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return self.text_[self.current:] 22601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 22701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def StartTranslateable(self): 22801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org assert not self.InTranslateable() 22901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if self.current != 0: 23001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Append a nontranslateable chunk 23101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org chunk_text = self.text_[self.chunk_start : self.last_nontranslateable + 1] 23201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Needed in the case where document starts with a translateable. 23301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if len(chunk_text) > 0: 23401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(False, chunk_text) 23501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunk_start = self.last_nontranslateable + 1 23601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_translateable = self.current 23701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = -1 23801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 23901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def EndTranslateable(self): 24001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org assert self.InTranslateable() 24101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Append a translateable chunk 24201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(True, 24301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.text_[self.chunk_start : self.last_translateable + 1]) 24401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunk_start = self.last_translateable + 1 24501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_translateable = -1 24601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = self.current 24701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 24801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def AdvancePast(self, match): 24901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.current += match.end() 25001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 25101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def AddChunk(self, translateable, text): 25201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Adds a chunk to self, removing linebreaks and duplicate whitespace 25301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if appropriate. 25401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''' 25501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _DESCRIPTION_COMMENT.search(text) 25601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 25701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_description = m.group('description') 2588a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # Remove the description from the output text 25901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org text = _DESCRIPTION_COMMENT.sub('', text) 26001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2618a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org m = _MESSAGE_BREAK_COMMENT.search(text) 2628a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org if m: 2638a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # Remove the coment from the output text. It should already effectively 2648a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # break apart messages. 2658a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = _MESSAGE_BREAK_COMMENT.sub('', text) 2668a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 2678a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org if translateable and not self.last_element_ in _PREFORMATTED_TAGS: 2688a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org if self.fold_whitespace_: 2698a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # Fold whitespace sequences if appropriate. This is optional because it 2708a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # alters the output strings. 2718a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = _FOLD_WHITESPACE.sub(' ', text) 2728a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org else: 2738a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = text.replace('\n', ' ') 2748a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = text.replace('\r', ' ') 2758a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # This whitespace folding doesn't work in all cases, thus the 2768a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org # fold_whitespace flag to support backwards compatibility. 2778a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = text.replace(' ', ' ') 2788a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text = text.replace(' ', ' ') 2798a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 28001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if translateable: 28101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org description = self.last_description 28201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_description = '' 28301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 28401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org description = '' 28501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 28601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if text != '': 28701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunks_.append((translateable, text, description)) 28801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2898a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org def Parse(self, text, fold_whitespace): 29001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Parses self.text_ into an intermediate format stored in self.chunks_ 29101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org which is translateable and nontranslateable chunks. Also returns 29201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunks_ 29301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 2948a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org Args: 2958a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org text: The HTML for parsing. 2968a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org fold_whitespace: Whether whitespace sequences should be folded into a 2978a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org single space. 2988a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 29901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Return: 30001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org [chunk1, chunk2, chunk3, ...] (instances of class Chunk) 30101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''' 30201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # 30301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Chunker state 30401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # 30501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 30601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.text_ = text 3078a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org self.fold_whitespace_ = fold_whitespace 30801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 30901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # A list of tuples (is_translateable, text) which represents the document 31001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # after chunking. 31101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunks_ = [] 31201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 31301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Start index of the last chunk, whether translateable or not 31401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunk_start = 0 31501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 31601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Index of the last for-sure translateable character if we are parsing 31701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # a translateable chunk, -1 to indicate we are not in a translateable chunk. 31801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # This is needed so that we don't include trailing whitespace in the 31901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # translateable chunk (whitespace is neutral). 32001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_translateable = -1 32101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 32201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Index of the last for-sure nontranslateable character if we are parsing 32301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # a nontranslateable chunk, -1 if we are not in a nontranslateable chunk. 32401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # This is needed to make sure we can group e.g. "<b>Hello</b> there" 32501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # together instead of just "Hello</b> there" which would be much worse 32601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # for translation. 32701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = -1 32801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 32901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Index of the character we're currently looking at. 33001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.current = 0 33101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 33201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # The name of the last block element parsed. 33301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_element_ = '' 33401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 33501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # The last explicit description we found. 33601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_description = '' 33701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 338807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org # Whether no-break was the last chunk seen 339807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org self.last_nobreak = False 340807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org 34101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org while self.current < len(self.text_): 34201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org _DebugPrint('REST: %s' % self.text_[self.current:self.current+60]) 34301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 344807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org m = _MESSAGE_NO_BREAK_COMMENT.match(self.Rest()) 345807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org if m: 346807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org self.AdvancePast(m) 347807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org self.last_nobreak = True 348807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org continue 349807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org 350807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org # Try to match whitespace 35101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _WHITESPACE.match(self.Rest()) 35201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 35301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Whitespace is neutral, it just advances 'current' and does not switch 35401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # between translateable/nontranslateable. If we are in a 35501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # nontranslateable section that extends to the current point, we extend 35601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # it to include the whitespace. If we are in a translateable section, 35701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # we do not extend it until we find 35801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # more translateable parts, because we never want a translateable chunk 35901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # to end with whitespace. 36001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if (not self.InTranslateable() and 36101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable == self.current - 1): 36201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = self.current + m.end() - 1 36301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AdvancePast(m) 36401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 36501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 36601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Then we try to match nontranslateables 36701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _NONTRANSLATEABLES.match(self.Rest()) 36801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 36901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if self.InTranslateable(): 37001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.EndTranslateable() 37101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = self.current + m.end() - 1 37201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AdvancePast(m) 37301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 37401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 37501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Now match all other HTML element tags (opening, closing, or empty, we 37601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # don't care). 37701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _ELEMENT.match(self.Rest()) 37801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 37901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org element_name = m.group('element').lower() 38001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if element_name in _BLOCK_TAGS: 38101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_element_ = element_name 38201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if self.InTranslateable(): 383807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org if self.last_nobreak: 384807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org self.last_nobreak = False 385807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org else: 386807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org self.EndTranslateable() 38701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 38801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Check for "special" elements, i.e. ones that have a translateable 38901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # attribute, and handle them correctly. Note that all of the 39001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # "special" elements are block tags, so no need to check for this 39101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # if the tag is not a block tag. 39201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org sm = _SPECIAL_ELEMENT.match(self.Rest()) 39301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if sm: 39401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Get the appropriate group name 39501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for group in sm.groupdict().keys(): 39601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if sm.groupdict()[group]: 39701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org break 39801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 39901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # First make a nontranslateable chunk up to and including the 40001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # quote before the translateable attribute value 40101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(False, self.text_[ 40201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunk_start : self.current + sm.start(group)]) 40301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Then a translateable for the translateable bit 40401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(True, self.Rest()[sm.start(group) : sm.end(group)]) 40501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Finally correct the data invariant for the parser 40601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.chunk_start = self.current + sm.end(group) 40701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 40801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_nontranslateable = self.current + m.end() - 1 40901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org elif self.InTranslateable(): 41001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # We're in a translateable and the tag is an inline tag, so we 41101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # need to include it in the translateable. 41201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_translateable = self.current + m.end() - 1 41301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AdvancePast(m) 41401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 41501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 41601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Anything else we find must be translateable, so we advance one character 41701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # at a time until one of the above matches. 41801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if not self.InTranslateable(): 41901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.StartTranslateable() 42001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 42101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.last_translateable = self.current 42201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.current += 1 42301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 42401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Close the final chunk 42501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if self.InTranslateable(): 42601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(True, self.text_[self.chunk_start : ]) 42701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 42801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.AddChunk(False, self.text_[self.chunk_start : ]) 42901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 43001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return self.chunks_ 43101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 43201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 43301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgdef HtmlToMessage(html, include_block_tags=False, description=''): 43401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Takes a bit of HTML, which must contain only "inline" HTML elements, 43501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org and changes it into a tclib.Message. This involves escaping any entities and 43601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org replacing any HTML code with placeholders. 43701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 43801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org If include_block_tags is true, no error will be given if block tags (e.g. 43901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org <p> or <br>) are included in the HTML. 44001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 44101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Args: 44201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org html: 'Hello <b>[USERNAME]</b>, how <i>are</i> you?' 44301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org include_block_tags: False 44401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 44501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Return: 44601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org tclib.Message('Hello START_BOLD1USERNAMEEND_BOLD, ' 44701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'howNBSPSTART_ITALICareEND_ITALIC you?', 44801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org [ Placeholder('START_BOLD', '<b>', ''), 44901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Placeholder('USERNAME', '[USERNAME]', ''), 45001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Placeholder('END_BOLD', '</b>', ''), 45101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Placeholder('START_ITALIC', '<i>', ''), 45201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Placeholder('END_ITALIC', '</i>', ''), ]) 45301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''' 45401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Approach is: 45501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # - first placeholderize, finding <elements>, [REPLACEABLES] and 45601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # - then escape all character entities in text in-between placeholders 45701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 45801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts = [] # List of strings (for text chunks) and tuples (ID, original) 45901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # for placeholders 46001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 46101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org count_names = {} # Map of base names to number of times used 46201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org end_names = {} # Map of base names to stack of end tags (for correct nesting) 46301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 46401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def MakeNameClosure(base, type = ''): 46501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Returns a closure that can be called once all names have been allocated 46601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org to return the final name of the placeholder. This allows us to minimally 46701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org number placeholders for non-overlap. 46801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 46901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Also ensures that END_XXX_Y placeholders have the same Y as the 47001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org corresponding BEGIN_XXX_Y placeholder when we have nested tags of the same 47101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org type. 47201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 47301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Args: 47401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org base: 'phname' 47501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org type: '' | 'begin' | 'end' 47601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 47701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Return: 47801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Closure() 47901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''' 480807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org name = base.upper() 48101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if type != '': 48201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org name = ('%s_%s' % (type, base)).upper() 48301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 48401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if name in count_names.keys(): 48501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org count_names[name] += 1 48601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 48701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org count_names[name] = 1 48801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 48901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def MakeFinalName(name_ = name, index = count_names[name] - 1): 49001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if (type.lower() == 'end' and 49101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org base in end_names.keys() and len(end_names[base])): 49201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return end_names[base].pop(-1) # For correct nesting 49301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if count_names[name_] != 1: 49401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org name_ = '%s_%s' % (name_, _SUFFIXES[index]) 49501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # We need to use a stack to ensure that the end-tag suffixes match 49601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # the begin-tag suffixes. Only needed when more than one tag of the 49701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # same type. 49801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if type == 'begin': 49901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org end_name = ('END_%s_%s' % (base, _SUFFIXES[index])).upper() 50001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if base in end_names.keys(): 50101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org end_names[base].append(end_name) 50201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 50301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org end_names[base] = [end_name] 50401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 50501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return name_ 50601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 50701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return MakeFinalName 50801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 50901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current = 0 510807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org last_nobreak = False 51101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 51201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org while current < len(html): 513807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org m = _MESSAGE_NO_BREAK_COMMENT.match(html[current:]) 514807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org if m: 515807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org last_nobreak = True 516807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org current += m.end() 517807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org continue 518807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org 51901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _NBSP.match(html[current:]) 52001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 52101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append((MakeNameClosure('SPACE'), m.group())) 52201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current += m.end() 52301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 52401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 52501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _REPLACEABLE.match(html[current:]) 52601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 52701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Replaceables allow - but placeholders don't, so replace - with _ 52801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ph_name = MakeNameClosure('X_%s_X' % m.group('name').replace('-', '_')) 52901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append((ph_name, m.group())) 53001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current += m.end() 53101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 53201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 53301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _SPECIAL_ELEMENT.match(html[current:]) 53401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 53501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if not include_block_tags: 536807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org if last_nobreak: 537807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org last_nobreak = False 538807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org else: 539807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org raise exception.BlockTagInTranslateableChunk(html) 54001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org element_name = 'block' # for simplification 54101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Get the appropriate group name 54201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for group in m.groupdict().keys(): 54301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m.groupdict()[group]: 54401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org break 54501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append((MakeNameClosure(element_name, 'begin'), 54601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org html[current : current + m.start(group)])) 54701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append(m.group(group)) 54801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append((MakeNameClosure(element_name, 'end'), 54901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org html[current + m.end(group) : current + m.end()])) 55001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current += m.end() 55101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 55201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 55301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _ELEMENT.match(html[current:]) 55401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 55501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org element_name = m.group('element').lower() 55601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if not include_block_tags and not element_name in _INLINE_TAGS: 557807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org if last_nobreak: 558807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org last_nobreak = False 559807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org else: 560807b78b2bb3a1ada6afb52bc0301ce717c230503joi@chromium.org raise exception.BlockTagInTranslateableChunk(html[current:]) 56101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if element_name in _HTML_PLACEHOLDER_NAMES: # use meaningful names 56201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org element_name = _HTML_PLACEHOLDER_NAMES[element_name] 56301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 56401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Make a name for the placeholder 56501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org type = '' 56601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if not m.group('empty'): 56701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m.group('closing'): 56801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org type = 'end' 56901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 57001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org type = 'begin' 57101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append((MakeNameClosure(element_name, type), m.group())) 57201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current += m.end() 57301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org continue 57401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 57501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if len(parts) and isinstance(parts[-1], types.StringTypes): 57601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts[-1] += html[current] 57701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 57801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org parts.append(html[current]) 57901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org current += 1 58001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 58101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg_text = '' 58201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org placeholders = [] 58301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for part in parts: 58401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if isinstance(part, types.TupleType): 58501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org final_name = part[0]() 58601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org original = part[1] 58701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg_text += final_name 58801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org placeholders.append(tclib.Placeholder(final_name, original, '(HTML code)')) 58901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 59001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg_text += part 59101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 59201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg = tclib.Message(text=msg_text, placeholders=placeholders, 59301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org description=description) 59401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org content = msg.GetContent() 59501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for ix in range(len(content)): 59601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if isinstance(content[ix], types.StringTypes): 59701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org content[ix] = util.UnescapeHtml(content[ix], replace_nbsp=False) 59801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 59901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return msg 60001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 60101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 60201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.orgclass TrHtml(interface.GathererBase): 60301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Represents a document or message in the template format used by 60401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Total Recall for HTML documents.''' 60501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 606ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org def __init__(self, *args, **kwargs): 607ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org super(TrHtml, self).__init__(*args, **kwargs) 60801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.have_parsed_ = False 60901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_ = [] # list of strings and MessageClique objects 6108a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org self.fold_whitespace_ = False 6118a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 6128a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org def SetAttributes(self, attrs): 6138a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org '''Sets node attributes used by the gatherer. 6148a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 6158a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org This checks the fold_whitespace attribute. 6168a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org 6178a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org Args: 6188a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org attrs: The mapping of node attributes. 6198a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org ''' 6208a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org self.fold_whitespace_ = ('fold_whitespace' in attrs and 6218a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org attrs['fold_whitespace'] == 'true') 62201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 62301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def GetText(self): 62401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Returns the original text of the HTML document''' 62501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return self.text_ 62601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 627ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org def GetTextualIds(self): 628ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org return [self.extkey] 629ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org 63001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def GetCliques(self): 63101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Returns the message cliques for each translateable message in the 63201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org document.''' 633705a118ab4a1f2fe348fccdcd4786a0b5bf426ecjoi@chromium.org return [x for x in self.skeleton_ if isinstance(x, clique.MessageClique)] 63401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 63501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def Translate(self, lang, pseudo_if_not_available=True, 63601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org skeleton_gatherer=None, fallback_to_english=False): 63701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org '''Returns this document with translateable messages filled with 63801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org the translation for language 'lang'. 63901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 64001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Args: 64101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org lang: 'en' 64201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org pseudo_if_not_available: True 64301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 64401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Return: 64501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 'ID_THIS_SECTION TYPE\n...BEGIN\n "Translated message"\n......\nEND 64601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 64701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org Raises: 64801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org grit.exception.NotReady() if used before Parse() has been successfully 64901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org called. 65001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org grit.exception.NoSuchTranslation() if 'pseudo_if_not_available' is false 65101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org and there is no translation for the requested language. 65201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org ''' 65301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if len(self.skeleton_) == 0: 65401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org raise exception.NotReady() 65501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 65601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # TODO(joi) Implement support for skeleton gatherers here. 65701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 65801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org out = [] 65901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for item in self.skeleton_: 66001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if isinstance(item, types.StringTypes): 66101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org out.append(item) 66201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 66301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg = item.MessageForLanguage(lang, 66401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org pseudo_if_not_available, 66501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org fallback_to_english) 66601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for content in msg.GetContent(): 66701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if isinstance(content, tclib.Placeholder): 66801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org out.append(content.GetOriginal()) 66901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 67001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # We escape " characters to increase the chance that attributes 67101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # will be properly escaped. 67201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org out.append(util.EscapeHtml(content, True)) 67301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 67401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return ''.join(out) 67501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 67601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org def Parse(self): 67701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if self.have_parsed_: 67801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org return 67901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.have_parsed_ = True 68001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 681b9161407f737461b5db16a29782f8a31d19e602dbenrg@chromium.org text = self._LoadInputFile() 682ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org 683ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org # Ignore the BOM character if the document starts with one. 684ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org if text.startswith(u'\ufeff'): 685ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org text = text[1:] 686ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org 687ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org self.text_ = text 688ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org 689ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org # Parsing is done in two phases: First, we break the document into 690ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org # translateable and nontranslateable chunks. Second, we run through each 691ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org # translateable chunk and insert placeholders for any HTML elements, 692ec8016c73b3b945b6284746230913d88653f35e7benrg@chromium.org # unescape escaped characters, etc. 69301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 69401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # First handle the silly little [!]-prefixed header because it's not 69501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # handled by our HTML parsers. 69601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org m = _SILLY_HEADER.match(text) 69701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if m: 69801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_.append(text[:m.start('title')]) 69901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_.append(self.uberclique.MakeClique( 70001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org tclib.Message(text=text[m.start('title'):m.end('title')]))) 70101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_.append(text[m.end('title') : m.end()]) 70201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org text = text[m.end():] 70301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 7048a303dc2cabedfd395b49f21310ea21c77056fe7joi@chromium.org chunks = HtmlChunks().Parse(text, self.fold_whitespace_) 70501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 70601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for chunk in chunks: 70701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if chunk[0]: # Chunk is translateable 70801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_.append(self.uberclique.MakeClique( 70901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org HtmlToMessage(chunk[1], description=chunk[2]))) 71001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org else: 71101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_.append(chunk[1]) 71201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 71301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # Go through the skeleton and change any messages that consist solely of 71401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org # placeholders and whitespace into nontranslateable strings. 71501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for ix in range(len(self.skeleton_)): 71601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org got_text = False 71701b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if isinstance(self.skeleton_[ix], clique.MessageClique): 71801b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org msg = self.skeleton_[ix].GetMessage() 71901b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org for item in msg.GetContent(): 72001b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if (isinstance(item, types.StringTypes) and _NON_WHITESPACE.search(item) 72101b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org and item != ' '): 72201b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org got_text = True 72301b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org break 72401b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org if not got_text: 72501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org self.skeleton_[ix] = msg.GetRealContent() 72601b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 72777cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org def SubstituteMessages(self, substituter): 72877cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org '''Applies substitutions to all messages in the tree. 72977cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org 73077cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org Goes through the skeleton and finds all MessageCliques. 73177cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org 73277cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org Args: 73377cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org substituter: a grit.util.Substituter object. 73477cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org ''' 73577cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org new_skel = [] 73677cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org for chunk in self.skeleton_: 73777cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org if isinstance(chunk, clique.MessageClique): 73877cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org old_message = chunk.GetMessage() 73977cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org new_message = substituter.SubstituteMessage(old_message) 74077cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org if new_message is not old_message: 74177cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org new_skel.append(self.uberclique.MakeClique(new_message)) 74277cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org continue 74377cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org new_skel.append(chunk) 74477cbaa8b1f1af05d8ba2c2a951c74e7909318830joi@chromium.org self.skeleton_ = new_skel 74501b3bc768461bd303bff39f8cd1663682254e407joi@chromium.org 746