pyVHDLParser
pyVHDLParser copied to clipboard
Paebbels
Subtype Constraint in Entity Port Causes Error
I'd like to help contribute to this project because I find there aren't enough VHDL parsers that are complete and open source, available for use in other projects.
This seems to be as good a place to start as any, so I cloned the repository and quickly found out the subtype constraint on an entity port doesn't pass the blockstreaming with the error ERROR: Expected ';', ':=' or whitespace after subtype indication.
To be clear, this happens whenever the port has a constraint, like a range constraint: s_axis_tdata : in std_logic_vector(31 downto 0);.
There are a lot of constraints which can apply to the subtype. I'm not necessarily worried about capturing them all, but I am interested in parsing out mostly entities and instantiations in code.
If you have a recommendation on where to look in the code to help add the capabilities, and a list of capabilities in order of precedence you want others to handle, I'd like to try to help out.
@bpadalino thanks for asked what can be done at pyVHDLParser.
The repository is still in an alpha state (v0.4.x) and contains some known errors:
- Broken doubled linked lists close to expressions.
This feature was added lately and still cause some problems. - Distinguishing between labels and assignments as well as procedure calls.
- Higher level stream processing (groups).
I think your example code is related to #6 .
Currently the parser has problems when the subtype indication isn't followed by :=. I'm currently investigating this issue.
Lately I added a lot of documentation (see https://pyvhdlparser.readthedocs.io/en/latest/). have you started to dig into some details? I wanted to add more details soon, so interested contributors can easily understand how it works and how to debug.
With regards to debugging I'm right now writing a functionality to export the internal data structures (linked lists / stream) to graphviz, so the parsed input can be displayed as SVG/PNG diagrams. Exporting tokens works fine, but exporting groups and tokens isn't done yet.
If you like, this could be a first task, to get more familiar with the internal data structures before fixing bugs in the parser itself.
I also started to upload testcases written with Pythons unittest. These are done for entities and architectures as well as tokens. The list of testcases could be expanded to:
- generics
- ports
- object declarations
- functions
- procedures
- packages
- package bodies
Tests based on unittest are written in Python itself, the aforementioned test had been started as *.vhdl files, which should be translated to unittests.
In general, I would like to get some help :)
That looks great! Where in the code is that at? I can try to take a look at adding this type of output for groups and tokens as you suggested.
As a note, I tried adding an initializer to my entity list, and it didn't seem to help the error.
I didn't notice much in the documentation on how the parser worked. For example, ERROR is raised here but it seems parserState is manipulated directly and you're creating little state machines for parsing each little part?
The issue with the parsing is probably because I am passing in (31 downto 0) and the parser wants the close of the entity ), the end of the statement ;, whitespace, or an assignment :=. I could add a new section to the FSM which reads the ( until there is a ), but that seems wrong. It would seem correct to make something that is a Constraint and look for different types of constraints. That might be an interesting section in the documentation to read how the parser is supposed to work and string those things together.
@bpadalino uploading an improved version, where you can study how I created the token stream diagram is currently blocked by Paebbels/pyAttributes#7. This fix is needed to slice the Frontend.py into multiple parts that can be handled more easily.
I also wrote down some notes on a paper, so I can explain to you how the parser and the states are working.
@bpadalino so Paebbels/pyAttributes#7 is now closed and I pushed a version of pyVHDLParser, where Frontend.py is split into multiple files to handle all the command line arguments.
next I'll write the docs how the ParserState data structure works.
As I pushed my updates, I found out, that ReadTheDocs isn't compiling my latest help pages, due to missing Python 3.8 support. I request an upgrade of my documentation repositories. I hope they do it soon. See https://github.com/readthedocs/readthedocs.org/issues/6324#issuecomment-581530254
Nonetheless, I'll write more documentation. I was requested to also document the planned CodeDOM.
