Lazy load dask.dataframe in datashader.py

[hoxbro]

Preparation for holoviz/datashader#1350

[codecov]

Codecov Report

Attention: Patch coverage is 94.44444% with 1 line in your changes missing coverage. Please review.

Project coverage is 88.50%. Comparing base (ccae8d1) to head (2c9d7af).
Report is 4 commits behind head on main.

Files	Patch %	Lines
holoviews/operation/datashader.py	80.00%	1 Missing ⚠️

Additional details and impacted files

@@           Coverage Diff           @@
##             main    #6309   +/-   ##
=======================================
  Coverage   88.49%   88.50%           
=======================================
  Files         323      323           
  Lines       68039    68093   +54     
=======================================
+ Hits        60210    60263   +53     
- Misses       7829     7830    +1     

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

[maximlt]

I'm fine with the current implementation but feel free to update it based on the comment.

[maximlt]

A bit nit but I tend to think this doesn't do what people mean with lazy import (i.e. delaying importing a module until it's used). I just remembered hvPlot has a few utils that do this together with the isinstance check, like:

def is_dask(data):
    if not check_library(data, 'dask'):
        return False
    import dask.dataframe as dd

    return isinstance(data, (dd.DataFrame, dd.Series))

def is_xarray_dataarray(data):
    if not check_library(data, 'xarray'):
        return False
    from xarray import DataArray

    return isinstance(data, DataArray)

[maximlt]

Just tried that in my env and got some weird results.

from holoviews.core.util import lazy_isinstance
import dask.dataframe as dd

ddf = dd.DataFrame({})

print(lazy_isinstance(ddf, 'dask.dataframe:DataFrame'))
print(isinstance(ddf, dd.DataFrame))

False
True

That's because obj.__module__ is dask_expr._collection in this case. Pretty unusual but well!

---

Alternative API with lazy_instance(obj, module, objname) where objname is either a string or a tuple of strings, e.g. lazy_instance(obj, 'dask.dataframe', 'DataFrame'), lazy_instance(obj, 'cudf', ('DataFrame', 'Series')).

[hoxbro]

I have changed it to .startswith. This is what is done in hvplot.