TileDB-Inc/TileDB-Py

Feature: Support `pathlib.Path`

Nicholas-Schaub opened this issue · 2 comments

Nicholas-Schaub commented

Summary

When reading/writing arrays, tiledb does not support pathlib.Path for file URIs.

Details

When using a pathlib.Path to open a new array, the first write triggers an array creation event that fails with the following error (truncated):

  File "tiledb/libtiledb.pyx", line 962, in tiledb.libtiledb.Array.create
  File "tiledb/libtiledb.pyx", line 527, in tiledb.libtiledb.unicode_path
  File "tiledb/libtiledb.pyx", line 520, in tiledb.libtiledb.ustring
TypeError: ustring() must be a string or a bytes-like object, not <class 'pathlib.PosixPath'>

What I Expect to Happen

When opening a tiledb.Array, I should be able to use a pathlib.Path object. Additionally, using the os pathing objects should also be possible.

Implementation

The stack trace seems to indicate the code is written in Cython. This should be a relatively easy feature to implement with two lines of code to type check if the input object is pathlib.Path, then casting with str(pathlib.Path("path/to/file.tiledb")).

If you're open to taking PRs, I am happy to contribute.

nguyenv commented

Hi @Nicholas-Schaub,

Thank you for reporting this issue. Yes, we are open to contributions! Please feel free to put in a PR.

Thanks.

thetorpedodog commented

The VFS change addresses this in part; I will probably expand essentially the same thing to the rest of the TileDB library in a follow-up.

  • 👍2