Changelogs

Version 2.1.4

• load now works as expected for ratexps in awalipy.
• In awalipy, aut_to_exp is now a method of class Automaton.
• In awalipy, explore is now accessible.
• In awalipy, aut_to_exp now works on ratexps.
• In awalipy, strip_history is now accessible.
• In awalipy, basic metric operations on ratexps are now accessible.
• In cora, minimal-automaton is now properly declared as possibly taking an automaton or an expression
• In cora, enumerate now longer display weights for Boolean automata
• Added an example automaton english-dict: it accepts a large but finite language.
• The CMake option CYTHON_BIN was not used, and it is clearly stated.
• Fixed several bugs in json printer.

Version 2.1.3

• Remove add_compile_definitions from CMakeLists.txt to enforce the compatibility of awali with CMake 3.1.
• The auto-join routine for contexts (e.g. with product) is now commutative and should no longer cause compilation errors of module promotion.
• Fixed a minor memory leak hapening when using legacy format fsm_json_v0
• In awalipy,
  • load now correctly loads example rational expressions.
  • aut-to-exp now is available for transducers
  • aut-to-exp now accepts a method optional argument as it already did in other layers.

Version 2.1.2

• explore_with_bound and explore_by_length are now correctly accessible from awalipy.
• minimal_automaton now treats correctly incomplete that are not co-accessible.
• The internal names of example automata are now always consistent with their loading names.

Version 2.1.1

• Fixed a major bug that would make awalipy fail at import on some configuration of MacOS.
• Fixed a fatal bug in Cora's command edit when the terminal was about 60 characters wide.
• Cora's command new now follows specification when used together with --output=json: it skips editting and prints an empty automaton.
• Cora's command new now sets correctly the name and caption of the automaton, when provided through options --caption/-C and --name/-N.

Version 2.1.0

• Added the functionality to explore part of the (weighted) determinisation of an automaton. It comes in two functions, which are available in all layers with the same name.
  • explore-by-length allows to bound by the length of words: the resulting automaton is correct for every word the length of which is below given depth.
  • explore-with-bound allows to bound each component of the weightsed subset-construction. This functionality requires the weightset to be either: ℕ, ℤ, ℕ∪∞, or N[k].
• Added the possibility to promote the weight semiring of rational expressions to super-semirings, just as it was already possible for automata.
  • In dyn, it is called promote_ratexp
  • In cora, it overloads the promote command
  • In python, it is a method of class RatExp.
• Added the function left_mult_standard in dyn.
  In cora, this function can be called through the command left-mult with option -Mstandard.
• Added the function subnormalize in dyn and cora. This function applies letterize_tape to the two first tapes of the automaton, followed by proper_here.
• The way awali parses numbers (integers, rationals, double) has been unified and sanitized across layers.
• In cora,
  • Merged commands eval-aut, and eval-word into eval. Command eval launch one of the three command depending on the type of parameter.
  • Added typecheck and smart typecast of input automata and transducers.
    • Cora now terminates with a fatal error if it should manipulate a transducer with three or more tapes.
    • Cora now issues a warning if it should do any meaningful transformation to an automaton or transducer in which labels are words. Although it is sometimes working, cora is not doing the necessary checks to safely apply algorithms to those types of automata/transducers.
    • An automaton that allows epsilon-transitions but contains none is now automatically transformed into an automaton that forbids epsilon-transitions. This behavior improves user experience. Previously, cora was quite often terminating with strange error messages where the system refuses to do harmless transformation due to typechecks.
    • Transducers are automatically transformed in such a way that, on each tape, epsilon is allowed as a label. In effect, this means that the only type of transducers allowed in cora is: two-tape transducers that allow epsilon on each tape.
• Added two CMake variables, AWALICPP_MODULE_COMPILE_DIR and COMPILE_MODULE_IN_INSTALL_DIR, to govern where are compiled the dyn modules at execution.

Changelogs for versions of Awali prior to

• Versions 2.0.X

Dernière Last modification: 11 Janvier January 2022