gTTS 2.0.0

[pndurette]

Hello!
I am super happy to announce the release of gTTS 2.0.0! This is pretty much a rewrite with ton of new cool stuff and fixes which I've been working on for the past few months. See below!
Also see new docs at http://gtts.readthedocs.io/

Upgrading to 2.0.0

Most users will have nothing to change!
See Deprecations and Removals below.

Features

• The gtts module
  • New logger ("gtts") replaces all occurrences of print()
  • Languages list is now obtained automatically (gtts.lang) (Supporting Languages #91, Filipino Language Support #94, when use "zh-yue" occured HTTPError: 404 Client Error #106)
  • Added a curated list of language sub-tags that have been observed to provide different dialects or accents (e.g. "en-gb", "fr-ca")
  • New gTTS() parameter lang_check to disable language checking.
  • gTTS() now delegates the text tokenizing to the API request methods (i.e. write_to_fp(), save()), allowing gTTS instances to be modified/reused
  • Rewrote tokenizing and added pre-processing (see below)
  • New gTTS() parameters pre_processor_funcs and tokenizer_func to configure pre-processing and tokenizing (or use a 3rd party tokenizer)
  • Error handling:
    • Added new exception gTTSError raised on API request errors. It attempts to guess what went wrong based on known information and observed behaviour ('NoneType' object has no attribute 'group' #60, when use "zh-yue" occured HTTPError: 404 Client Error #106)
    • gTTS.write_to_fp() and gTTS.save() also raise gTTSError on gtts_token error
    • gTTS.write_to_fp() raises TypeError when fp is not a file-like object or one that doesn't take bytes
    • gTTS() raises ValueError on unsupported languages (and lang_check is True)
    • More fine-grained error handling throughout (e.g.request failed vs. request successful with a bad response)
• Tokenizer (and new pre-processors):
  • Rewrote and greatly expanded tokenizer (gtts.tokenizer)
  • Smarter token 'cleaning' that will remove tokens that only contain characters that can't be spoken (i.e. punctuation and whitespace)
  • Decoupled token minimizing from tokenizing, making the latter usable in other contexts
  • New flexible speech-centric text pre-processing
  • New flexible full-featured regex-based tokenizer (gtts.tokenizer.core.Tokenizer)
  • New RegexBuilder, PreProcessorRegex and PreProcessorSub classes to make writing regex-powered text pre-processors and tokenizer cases easier
  • Pre-processors:
    • Re-form words cut by end-of-line hyphens
    • Remove periods after a (customizable) list of known abbreviations (e.g. "jr", "sr", "dr") that can be spoken the same without a period
    • Perform speech corrections by doing word-for-word replacements from a (customizable) list of tuples
  • Tokenizing:
    • Keep punctuation that modify the inflection of speech (e.g. "?", "!")
    • Don't split in the middle of numbers (e.g. "10.5", "20,000,000") (Punctuation is wrong with longer text  #101)
    • Don't split on "dotted" abbreviations and accronyms (e.g. "U.S.A")
    • Added Chinese comma ("，"), ellipsis ("…") to punctuation list to tokenize on (Missing punctuation #86)
• The gtts-cli command-line tool
  • Rewrote cli as first-class citizen module (gtts.cli), powered by Click
  • Windows support using setuptool's entry_points
  • Better support for Unicode I/O in Python 2
  • All arguments are now pre-validated
  • New --nocheck flag to skip language pre-checking
  • New --all flag to list all available languages
  • Either the --file option or the <text> argument can be set to "-" to read from stdin
  • The --debug flag uses logging and doesn't pollute stdout anymore

Bugfixes

• _minimize(): Fixed an infinite recursion loop that would occur when a token started with the miminizing delimiter (i.e. a space) (Missing punctuation #86)
• _minimize(): Handle the case where a token of more than 100 characters did not contain a space (e.g. in Chinese).
• Fixed an issue that fused multiline text together if the total number of characters was less than 100
• Fixed gtts-cli Unicode errors in Python 2.7 (famous last words) (Fixed unicode decoding issue. #78, codec can't decode byte 0xe3 in position 0: ordinal not in range(128) #93, Can't make mp3 file from the text in Romanian  #96)

Deprecations and Removals

• Dropped Python 3.3 support
• Removed debug parameter of gTTS (in favour of logger)
• gtts-cli: Changed long option name of -o to --output instead of --destination
• gTTS() will raise a ValueError rather than an AssertionError on unsupported language

Improved Documentation

• Rewrote all documentation files as reStructuredText
• Comprehensive documentation writen for Sphinx, published to http://gtts.readthedocs.io
• Changelog built with towncrier

Misc

• Major test re-work
• Language tests can read a TEST_LANGS enviromment variable so not all language tests are run every time.
• Added AppVeyor CI for Windows
• PEP 8 compliance