Improve SqlTypeName to support more types and also improve error handling

[andygrove]

• Improve SqlTypeName::from_string() to support more types, using the sql parser to parse parameterized types such as VARCHAR(n) and DECIMAL(p, s)
• Replaced some todo! and unimplemented! with Err and made corresponding changes in call sites

[codecov-commenter]

Codecov Report

Merging #824 (409a09b) into main (a7583b5) will decrease coverage by 0.02%.
The diff coverage is n/a.

@@            Coverage Diff             @@
##             main     #824      +/-   ##
==========================================
- Coverage   74.88%   74.86%   -0.03%     
==========================================
  Files          71       71              
  Lines        3588     3588              
  Branches      748      748              
==========================================
- Hits         2687     2686       -1     
+ Misses        771      768       -3     
- Partials      130      134       +4     

Impacted Files	Coverage Δ
dask_sql/_version.py	32.27% <0.00%> (-0.29%)	⬇️

📣 We’re building smart automated test selection to slash your CI/CD build times. Learn more

[randerzander]

I've just tried these changes with my own tests.

I can now use "string" types, but varchars w/ a defined limit still fail:

Traceback (most recent call last):               
  File "<stdin>", line 1, in <module>                                                                            
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/context.py", line 238, in create_table
    dc = InputUtil.to_dc(                                                                                                   
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/input_utils/convert.py", line 68, in to_dc
    table = filled_get_dask_dataframe(input_item)                                                           
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/input_utils/convert.py", line 57, in <lambda>
    filled_get_dask_dataframe = lambda *args: cls._get_dask_dataframe(                                                      
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/input_utils/convert.py", line 90, in _get_dask_dataframe
    return plugin.to_dc(                                                                               
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/input_utils/hive.py", line 69, in to_dc
    column_information = {                                                                                    
  File "/opt/conda/envs/rapids/lib/python3.9/site-packages/dask_sql/input_utils/hive.py", line 70, in <dictcomp>
    col: sql_to_python_type(SqlTypeName.fromString(col_type.upper()))                                      
RuntimeError: Internal("Cannot determine SQL type name for 'VARCHAR(65535)'")

[andygrove]

I have now pushed changes to support VARCHAR(n) and some other parameterized types

[andygrove]

cargo test is failing with:

  /usr/local/cargo/registry/src/github.com-1ecc6299db9ec823/pyo3-0.17.1/src/types/any.rs:813: undefined reference to `PyObject_Str'
          collect2: error: ld returned 1 exit status

[andygrove]

cargo test fails on this PR but not on others, so something in this PR is causing a linker issue. Very odd. I am looking into it.

[andygrove]

TIL that you cannot reference PyMethods from Rust unit tests (at least, not without additional environment changes). Tests should now be passing.