[foreign] RFC: Additional jextract filtering

[Jorn Vernee]

I see what you mean. Internal symbols might leak into the output as 
well.

Filtering by folder seems a little ad-hoc to me. Maybe what would work 
better is to only include things from the direct header files passed to 
jextract, and then allow users to specify a folder of headers as well 
using some wild-card pattern.

Then we can use the root-set to throw an error when something that is 
required is dropped. The user can either specify that header as well, or 
otherwise generate a separate jextract artifact of the dependency and 
put that on the class path when running jextract, and eventually 
runtime.

Jorn

Maurizio Cimadamore schreef op 2019-03-11 16:29:
> Btw, as for macros and enums, I'm afraid the current logic is not
> minimal enough; for instance, it is very likely that almost all
> headers would pull in constants from common headers such as stdint.h -
> do we really want those to show up? Most of the times those constants
> are relevant for the implementation of the library, much much less for
> the library client.
> 
> I'd propose that for enums and macros we'd adopt a path-based
> filtering, so that we do not add macro symbols and/or enums for
> headers that are in a different folder from the one that is the target
> of jextract.
> 
> (of course there should be a way to disable this adaptive filtering -
> which is probably needed for system libs - but my guess is that in 99%
> of cases it will just works for 3rd party libs such as python,
> tensorflow, sqlite & co).
> 
> Maurizio
> 
> On 11/03/2019 15:12, Maurizio Cimadamore wrote:
>> 
>> On 11/03/2019 13:45, Jorn Vernee wrote:
>>> I can separate the parts of the patch a little bit into; Filter 
>>> refactor + root set compute, and then leave the option changes out of 
>>> it. But those 2 alone do not affect the filtering, since the root set 
>>> is only used when filtering non-symbol/macro elements.
>> 
>> I guess then what I'm suggesting is to automatically filter out 
>> elements not in the root set, and see how that works out.
>> 
>> Maurizio
>> 
>>