updates to grammar to successfully generate lexer/parser/visitor with Antlr4.7.1

[clinssen]

No description provided.

[clinssen]

OK, so this is not exactly the latest version of Antlr4, but the one from the Ubuntu package repository--the actually latest Antlr4 (4.7.1) fails to parse the grammar, and complains about "custom channels are not supported in combined grammars". This might be a separate issue that we have to figure out.

[clinssen]

A slight change in the Antlr4 grammar (PyNestML.g4) seems to have fixed the issue.

[clinssen]

I filed this PR because I couldn't generate the lexer/parser/visitor.py from the PyNESTML grammar with antlr4.7.1.

In general, Antlr recommends that you separate your lexer grammar and your parser grammar. The lexer grammar is the old Tokens.g4, but another (new?) requirement of Antlr is that you explicitly define all your literals. That is, you cannot use '=' in the parser, but have to use, say, EQUALS, with EQUALS : '=' defined in the lexer. See e.g. antlr/antlr4#37

This PR implements the necessary changes. Note that the target language is set to Python2; this is also Python3 compatible (but not vice versa).

I would still like to include lexer/parser/visitor generation as part of the Travis CI build script.

[clinssen]

With the latest commit, lexer/parser/visitor generation has been included in the Travis CI build script. This means that during installation, the pynestml/generated directory is cleared, and antlr4 is invoked to generate lexer/parser/visitor.py, which are then used by pynestml in the further tests.

[jougs]

Many thanks for this update. I did not check any single line of the change and trust that @clinssen knows what he is doing.