[ARROW-9401] [C++/Python] Support necessary functionality to have an Arrow-string type in pandas

• Type: Wish
• Status: Open
• Priority: Major
• Resolution: Unresolved
• Component/s: C++, Python
• External issue URL: https://github.com/apache/arrow/issues/25481
• Assignee: Unassigned
• Reporter: Uwe Korn
• Created: 10/Jul/20 09:27
• Updated: 11/Jan/23 08:06

Description

This should serve as an umbrella issue for the needed functionality to have an Apache Arrow backed string type in pandas. In addition to the string kernels (ARROW-555), we probably need to implement some more support functionality to efficiently the pandas interfaces.

Some of these functions are already present in fletcher but a native string type in pandas should not have a hard dependency on numba.

Issue Links

• depends upon: ARROW-555 [C++] String algorithm library for StringArray/BinaryArray
• depends upon: ARROW-9431 [C++/Python] Kernel for SetItem(IntegerArray, values)
• depends upon: ARROW-9432 [C++/Python] Add option to Take kernel to interpret negative indices as indexing from the right
• depends upon: ARROW-9433 [C++/Python] Add option to Take kernel to interpret negative indices as NULL
• depends upon: ARROW-9407 [Python] Accept pd.NA as missing value in array constructor
• depends upon: ARROW-9429 [Python] ChunkedArray.to_numpy
• depends upon: ARROW-9430 [C++/Python] Kernel for SetItem(BooleanArray, values)
• depends upon: ARROW-9489 [C++] Add fill_null kernel implementation for (array[string], scalar[string])