search

stfc / fparser

This project maintains and develops a Fortran parser called fparser2 written purely in Python which supports Fortran 2003 and some Fortran 2008. A legacy parser fparser1 is also available but is not supported. The parsers were originally part of the f2py project by Pearu Peterson.
https://fparser.readthedocs.io
Other
61 stars 29 forks source link

allow fparser2 list classes to contain a single class #170

Closed rupertford closed 5 years ago

rupertford commented 5 years ago

fparser2 generates a list class whenever there is a Fortran rule that specifies a list. In the current implementation a list object is only created if there are two or more items in the list. If there is only one item then the item itself is returned, rather than a list containing the item.

As an example, consider the Fortran2003 rule for a calling a subroutine and its arguments ...

R1218 call-stmt is CALL procedure-designator [ ( [ actual-arg-spec-list ] ) ]
R1220 actual-arg-spec is [ keyword = ] actual-arg

If there are two arguments then we will get the following hierarchy

Call_Stmt
    Actual_Arg_Spec_List
        Actual_Arg
        Actual_Arg

but if there is only one we will get

Call_Stmt
    Actual_Arg

There is nothing wrong with this is terms of correctness. The problem comes when trying to traverse the parse tree. If you want to get access to all Arguments then you need to add in extra code to deal with the case when there is only one.

This is the situation for all generated lists in fparser2 and there are many of them (just search for *-list in the spec).

The proposed solution is to have a list object created when there are 1 or more items in the list. I don't think there is any fundamental reason why this can't be done in fparser2 and on face value it seems quite a simple change but it would nee to be investigated.

This would mean a change to the generated fparser2 parse tree which would affect anyone who has written code that relies on the current structure but I think it is worth it.

pelson commented 5 years ago

Looks like this is as simple as changing one number in https://github.com/stfc/fparser/blob/0.0.8/src/fparser/two/utils.py#L620.

In the work that I've done with path based querying for things like a UML generator, I've got an interface to stuff like for actual_arg in node.find_decendants('Actual_Arg'): .... In that situation, it is transparent to the user whether there was a list or not. Given this, I could see an argument against making all list-able things be lists even if they are length 1. I personally don't have a preference either way.

rupertford commented 5 years ago

Created branch 170_single_object_lists

arporter commented 5 years ago

203 has been merged. Closing.