Read CSV format text from stdin or memory

[heymind]

Migrate from apache/arrow#10066.

Which issue does this PR close?

Close Jira issue ARROW-12306. Closes #198

Rationale for this change

Let CsvFile and CsvExec support reading from a reader (not only files)

What changes are included in this PR?

This pr adds the following new pub functions:

• CsvFile::try_new_from_reader
• CsvFile::try_new_from_reader_infer_schema
• CsvExec::try_new_from_reader
• CsvStream::try_new_from_reader

Are there any user-facing changes?

No.

[alamb]

I think this PR is good -- the only thing that is wonky to me is that if you Clone one of these readers it won't work. However, in this case the error will be clear so I think this PR can be merged and if anyone cares about the Clone behavior we can fix it in a follow on PR.

Thanks again @heymind

What do you think @Dandandan / @andygrove ?

[codecov-commenter]

Codecov Report

Merging #54 (10f8c3b) into master (9ba214a) will decrease coverage by 0.04%.
The diff coverage is 73.29%.

❗ Current head 10f8c3b differs from pull request most recent head c938270. Consider uploading reports for the commit c938270 to get more accurate results

@@            Coverage Diff             @@
##           master      #54      +/-   ##
==========================================
- Coverage   76.24%   76.20%   -0.05%     
==========================================
  Files         134      134              
  Lines       23051    23199     +148     
==========================================
+ Hits        17576    17679     +103     
- Misses       5475     5520      +45     

Impacted Files	Coverage Δ
datafusion/src/datasource/csv.rs	73.33% <69.86%> (-9.28%)	⬇️
datafusion/src/physical_plan/csv.rs	79.09% <75.72%> (-4.13%)	⬇️

---

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 9ba214a...c938270. Read the comment docs.

[Dandandan]

How would this look if we would support multiple readers/partitions?

[Dandandan]

Also I am wondering, if we do it like this, we would we need to do the same for json/xml/etc sources that should support essentially the same to avoid reimplementing it for each format.

[alamb]

I suggest we file a follow on ticket for this work (supporting Reader input for JSON / XML sources). As you say I don't think it needs to be part of this PR.

[Dandandan]

Looks good to me, thanks @heymind .
Seems like a cool addition to integrate with much more sources. I have some small questions about partitioning and re-use across formats, but that shouldn't be a blocker.