v0.3.0-rust

• Add accept_bytes and token_ids_to_finish methods to EngineLike and implement them in Engine.

v0.2.1-python

• Changes from v0.3.0-rust.

v0.2.0-rust

• Make duplicate tokens in vocabulary initialization no longer a hard error but a log warning.
• Support __copy__ in Engine in Python.
• Check tokens' first bytes to warn against potentially problematic vocabulary.

v0.2.0-python

• Changes from v0.2.0-rust

v0.1.12-rust

• Remove with_config in Python FFI code and put its functionality into Python FFI constructor.

v0.1.8-python

• Engine constructor now accepts optional config parameter.
• Changes in v0.1.12-rust

v0.1.11-rust

• Fixed a bug that overwrites kbnf start nonterminal
• Fixed a bug that leads to out of bound index in Grammar debug method

v0.1.10-rust

• Removed debug println! from the library that are accidentally left by v0.1.9-rust

v0.1.9-rust

• Fixed a bug that makes first bytes filter inefficient
• Added anchors to regex strings so the DFA will immediately enter dead state when a byte that definitely ends a match is received by the DFA. This ensures the DFA will not accept one more arbitrary byte at the end of match, which is proven to be catastrophical.

v0.1.8-rust

• Fixed a bug: first bytes filter are now computed correctly for earley items on non-initial state. Previously, they assume all earley items are on initial state, which is very incorrect.