feat(python): Add read_clipboard and DataFrame.write_clipboard

[CanglongCl]

Closes #9902

Similar to pandas.DataFrame.to_clipboard and pandas.read_clipboard, read_clipboard reads csv format text from clipboard and convert it into DataFrame whereas DataFrame.write_clipboard writes DataFrame into clipboard with csv format.

These 2 methods are useful while exploring data from Excel or other similar software in interactive environment like jupyter notebook.

Implementation

• Reading and writing clipboard uses arboard in rust side, supporting Windows, macOS and Linux, maintained by 1Password.
• read_clipboard just reads the clipboard then passes the result to read_csv with default separator '\t'.
• write_clipboard just gets csv string by calling DataFrame.write_csv (but with default separator '\t') and writes to clipboard.

Difference from pandas method

pandas.read_clipboard uses regex '\\s+' as separator, but polars does not allow regex as separator in read_csv.

Compatibility

Tested Compatible:

• Microsoft Excel (Mac)
• Google Sheet (Online)
• WPS (Online)

Others

Both methods are only tested on M1 macOS. It would be much appreciated if someone could test on Linux and Windows.

[codecov]

Codecov Report

Attention: Patch coverage is 42.50000% with 23 lines in your changes are missing coverage. Please review.

Project coverage is 81.34%. Comparing base (252702a) to head (56002cc).
Report is 14 commits behind head on main.

Files	Patch %	Lines
py-polars/src/functions/io.rs	0.00%	18 Missing ⚠️
py-polars/polars/io/clipboard.py	72.72%	3 Missing ⚠️
py-polars/polars/dataframe/frame.py	50.00%	2 Missing ⚠️

Additional details and impacted files

@@            Coverage Diff             @@
##             main   #15272      +/-   ##
==========================================
+ Coverage   81.32%   81.34%   +0.01%     
==========================================
  Files        1359     1365       +6     
  Lines      176076   176655     +579     
  Branches     2526     2526              
==========================================
+ Hits       143191   143694     +503     
- Misses      32402    32478      +76     
  Partials      483      483              

☔ View full report in Codecov by Sentry.
📢 Have feedback on the report? Share it here.

[ritchie46]

I am not convinced by this one. Two things I am concerned about.

• New dependencies and binary size. (Can you check how much this increments by the new dependencies)
• Why default to tab separated data? What do we think about limited dtype support?

[CanglongCl]

I am not convinced by this one. Two things I am concerned about.

• New dependencies and binary size. (Can you check how much this increments by the new dependencies)
• Why default to tab separated data? What do we think about limited dtype support?

I'm not kinda familiar with this. I hope I did the right thing. I use my mac and make build-release in ./py-polars; and then I check the binary size of py-polars/polars/polars.abi3.so

• Before (252702a): 69.3 MB (69,303,536 Bytes)
• After (56002cc): 69.3 MB (69,306,240 Bytes)
• Difference: +2.704 KB

About default separator, while copying data in Excel, Google Sheet or WPS, the data will be transformed into csv with separator \t. Since these 2 methods are majorly used for exploring data in these softwares, \t is used as default. Similarly, pandas also does the same thing.

[cjackal]

• Why default to tab separated data?

Adding to @CanglongCl 's, rendered html table is another source of clipboard I/O which defaults to tab separator. "I/O of tsv strings via clipboard" is the force of nature, I think.

[ritchie46]

Alright. Let's give this a try. I understand why this can be nice. Thank you @CanglongCl