FIX-#7039: Pass scalar dtype as is to astype query compiler

[arunjose696]

What do these changes do?

Currently in master, In astype method if the datatype passed is a scalar, the scalar is converted to a dictionary with the keys as column names and values the scalar dtype value for all keys. This causes the exception message to be different in pandas vs modin as the function passed to remote function applies astype with the dictionary instead of scalar.

This PR passes the scalar dtypes as is to QC layer and the scalar is used in remote function for astype. This would also avoid unnecessarily iterating over the column names if dtype is a scalar value.

• first commit message and PR title follow format outlined here

  NOTE: If you edit the PR title to match this format, you need to add another commit (even if it's empty) or amend your last commit for the CI job that checks the PR title to pick up the new PR title.

• passes flake8 modin/ asv_bench/benchmarks scripts/doc_checker.py
• passes black --check modin/ asv_bench/benchmarks scripts/doc_checker.py
• signed commit with git commit -s
• Resolves test_series.py::test_astype failed due to different exception messages #7039?
• tests added and passing
• module layout described at docs/development/architecture.rst is up-to-date

[arunjose696]

Performance comparison for astype with scalar dtype

import modin.pandas as pd
import numpy as np
import string
from modin.utils import execute
import time
# Define the number of columns and rows
num_columns = 100000
num_rows = 5
data = {}
for i in range(num_columns):
    data[f'col_{i}'] = np.random.randint(0, 100, num_rows)

# Create DataFrame
df = pd.DataFrame(data)
execute(df)
start = time.time()
df.astype("str")
execute(df)
end = time.time()
print(end-start)

For Master 4.155833005905151 seconds
For Current 0.03567790985107422 seconds

[anmyachev]

Is it safe to use the same object in multiple elements of a series?

[arunjose696]

As it would be the same value for dtype for all elements of the series, What would be an issue using same object?

[anmyachev]

I believe we store a reference to the same object in several elements of the series, and if so, then changing one element will entail changing all elements, which can easily lead to errors. However, if this is not the case, then there is no problem.

[arunjose696]

As we only store dtype values to new_dtypes[:] I think it should be okay as the cases where values of dtypes itself being mutated inplace are very unlikely. If we reassign the value for dtypes eg with new_dtypes[0]="str" this would also work fine so I think we can keep this.

[anmyachev]

Considering also that this class does not have inplace operations other than materialization, then ok.

[anmyachev]

This argument can only be a string, right?

[arunjose696]

Yes this should take only string have made, have made changes to use apply() function and pass index of the column as a string.

[YarShev]

CI failed, please recheck.

[YarShev]

CI is failing because of #7167 so let's wait for it to be merged.

[YarShev]

@anmyachev, any comments?

[YarShev]

@arunjose696, please rebase on current master to make CI pass.

[YarShev]

rebased

[YarShev]

@anmyachev, any comments?

[anmyachev]

LGTM!