[erlang-questions] soft/optional typing
Wed Jun 13 10:14:54 CEST 2007
Raoul Duke wrote:
> What with HiPE and Dialyzer, seems like optional/soft/pluggable typing
> in Erlang (ideally with some inference) isn't such an odd thing.
> Googling around it appears there have been various developments along
> those lines - is there anything currently active?
We have a type inference for erlang based of the notion of Success
Typings described in
http://user.it.uu.se/~tobiasl/publications/succ_types.pdf This type
inference is the basis of Dialyzer's plt analysis. With the upcoming
release of Erlang/OTP R11B-5 you can access the full analysis by giving
the option --succ_typings to dialyzer. Many more bugs are found, but
some of the warnings might be hard to interpret. This is the reason it
is not on by default, but we are working to weed these problems out.
Based on the same analysis is also the type annotation tool Typer
(currently an unsupported part of otp). You can look at the description
in http://user.it.uu.se/~tobiasl/publications/typer.pdf although the
current version differs a bit from the description. Try bin/typer -h for
some more instructions on how to use it.
A current work is to incorporate a contract language in the framework,
in both order to get better analysis results, and to serve as a
complement to the edoc documentation. These contracts are currently
expressed as type schemes with bounded quantification. The first
priority is not to build a type checker since this would limit the
expressibility of Erlang. Instead we would like the users to
incrementally add type information where they want, and the anlysis can
tell you where the information cannot be correct. If the analysis cannot
find any inconsistencies, the contracts are trusted and can be used to
find better information in the later stages.
The results from the contract work looks promising and we hope to
present it on the Erlang Workshop.
> many thanks.
> erlang-questions mailing list
More information about the erlang-questions