allow to read non-standard CSV

[kazuk]

Which issue does this PR close?

Closes #315.

Rationale for this change

What changes are included in this PR?

Reader::from_reader split into Reader::build_csv_reader, Reader::from_csv_reader.

Reader::build_csv_reader builds csv_crate::Reader,with all CSV reader options.
Rader::from_csv_reader builds arrow::csv::reader::Reader for csv_crate::Reader.

add fn infer_file_schema_with_csv_options .

change infer_file_schema calls infer_file_schema_with_csv_options with default options.

add fn infer_reader_schema_with_csv_options.
change infer_reader_schema calls infer_reader_schema with default options.

add escape quote terminator to ReaderBuilder

ReaderBuilder::build uses added options.

Are there any user-facing changes?

currently minimized API change.

• add fn ReaderBuilder::escape
• add fn ReaderBuilder::quote
• add fn ReaderBuilder::terminator

please concider make public for fn infer_file_schema_with_csv_options, infer_reader_schema_with_csv_options , Reader::build_csv_reader , Reader::from_csv_reader.

[alamb]

Thank you for the contribution @kazuk

This code looks really nice and clean. 👍

The only thing that I think is missing from this PR is some basic tests -- specifically to show everything hooked together properly and the options get to the csv reader.

I think such tests would be of most value to ensure that as we change this code in the future, we don't accidentally break this new functionality. Give we are delegating to the implementation in the csv reader crate I don't think we need to test a large number of corner cases -- just that we can read a CSV file with a non default escape, quote and terminator character

Thanks again!

[codecov-commenter]

Codecov Report

Merging #326 (4bf2e0c) into master (c863a2c) will decrease coverage by 0.02%.
The diff coverage is 66.66%.

@@            Coverage Diff             @@
##           master     #326      +/-   ##
==========================================
- Coverage   82.52%   82.50%   -0.03%     
==========================================
  Files         162      162              
  Lines       44007    44036      +29     
==========================================
+ Hits        36316    36331      +15     
- Misses       7691     7705      +14     

Impacted Files	Coverage Δ
arrow/src/csv/reader.rs	86.97% <66.66%> (-1.39%)	⬇️
arrow/src/array/transform/boolean.rs	76.92% <0.00%> (-7.70%)	⬇️
parquet/src/encodings/encoding.rs	94.85% <0.00%> (-0.20%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update c863a2c...4bf2e0c. Read the comment docs.

[kazuk]

Thank you for review.

The only thing that I think is missing from this PR is some basic tests -- specifically to show everything hooked together properly and the options get to the csv reader.

I add a basic tests for detects feature broken.

[alamb]

Looks great to me -- thank you @kazuk

[alamb]

The MIRI CI check is not related to this PR (edit #345)

However, the lint failure seems to be: https://github.com/apache/arrow-rs/pull/326/checks?check_run_id=2658268180

I think it can be resolved by running cargo fmt locally and then pushing the result.

[kazuk]

oh!, Thanks.

I run cargo fmt and push.

[alamb]

The MIRI failure is unrelated to this PR: #345

[alamb]

Thanks again @kazuk