Add server computation

[MarcSkovMadsen]

Continues from #15, so review and merge that one first.

Adds server computation:

Reference Links:

• Graphic Walker Data Computation API
• Graphic Walker DSL Parser
• Graphic Walker DSL Parser for Python
• PygWalker: payload_to_sql

Todo

• [x] Philipp: Implement robust request-response communication for server based computations
  • [x] Test that communication system robust for multiple GraphicWalker instances in same app and when having multiple sessions/ users.
• [x] Make sure the object/ data does not get sent to client when server_computation=True.
• [x] Update PY.CAFE server computation example + App Demo example
• [x] Philipp: Fix export method.
• [x] Philipp: Resolve comments
• [ ] Philipp: Final tests and then release new version

maybe now/ maybe later

• [x] Fix as many #Todo items as possible.
• [x] Refactor js code to make it more readable and maintainable

Test Script

You can test the functionality by serving the code below.

import pandas as pd
import panel as pn

from panel_gwalker import GraphicWalker
import logging

pn.extension()

df = pd.read_parquet(
    "https://datasets.holoviz.org/windturbines/v1/windturbines.parq"
)
# Enable server-side computation for scalable data processing
walker = GraphicWalker(df, server_computation=True, _debug=True)

pn.Column(
    walker,
    walker.param.server_computation,
).servable()

[MarcSkovMadsen]

Got a POC working

https://github.com/user-attachments/assets/202e9d61-c592-42e1-97e9-cd0f13ad5739

[MarcSkovMadsen]

Will wait for help from @philippjfr. See https://github.com/philippjfr/panel-graphic-walker/pull/17#discussion_r1818149676.

[MarcSkovMadsen]

Morning @philippjfr . I've cleaned up the code and documentation. I've also added a theme parameter.

I'll stop here. I hope you will finalize. See todo list in first post.

Next step for me would probably be to add the TableWalker it can be a pretty good stand alone table supplementing Tabulator and Perspective. Then support for showing an initial chart in GraphicWalker. Then support for DuckDB table object (and Polars DataFrame) would be nice.