[EPIC] Streaming partitioned writes

[alamb]

Is your feature request related to a problem or challenge?

This is a tracking epic for a collection of features related to writing data.

The basic idea is better / full support for writing data:

1. to multiple (possibly Partitoned by value) files
2. To different file types (parquet, csv, json, avro, arrow)
3. In a streaming fashion (input doesn't need to be entirely buffered)
4. From SQL (via INSERT, INSERT INTO, COPY, Etc)
5. Stream to a target object_store (aka multi-part S3 upload)

This is partially supported today programmatically (see SessionContext::write_csv, etc)

Subtasks:

• Using INSERT INTO query to append a file #5130
• Add support for appending data to external tables - CSV #6526
• Implement COPY ... TO statement  #5654
• End to end SQL tests for writing data to a file (e.g for Add support for appending data to external tables - CSV #6526)
• Tests for streaming writes (that they actually stream while writing rather than buffering)
• Tests for abort behavior (making sure all writers are canceled and the correct error is returned)
• Implement the write API for JSON and parquet files (Unify DataFrame and SQL (Insert Into) Write Methods #7141 and Extend insert into to support Parquet backed tables #7244)
• Update the user guide Documentation to say that DataFusion supports streaming multi-part writes
• Re-implement DataFrame.write_* to use LogicalPlan::Write #5076
• [DataFrame] Parallel Write out of dataframe  #7079
• Allowing to expose sort order of parquet file to datafusion does not work #7036
• Implement support for FileFormat Options for COPY and Create External Table #7298
• Allowing setting sort order of parquet files without specifying the schema #7317
• Document DML statements #7319
• Create CopyOptions for controlling copy behavior #7322
• Allow inserts to a sorted ListingTable #7354
• Allow Setting Column Specific Parquet Write Options via SQL StatementOptions #7442
• Document COPY parquet specific options  #7499
• CSV/JSON Writes Not Guarenteed to Preserve Expected Ordering #7536
• Enable allow_single_file_parallelism by default to write out parquet files in parallel #7590
• Create page and column statistics when a parquet file is written in parallel #7589
• Limit parallelism / buffering when writing to a single parquet file in parallel  #7591
• Support AVRO Format for Write Queries #7679
• Allow Inserts to Partitioned Listing Table #7744
• [DataFrame] Parallel Load into dataframe #6983
• Dynamically Determine the Number of Output Files based on Configs #7767
• The output of write_csv and write_json methods results in many empty output files, which is confusing #5383
• Error writing to a partitioned table: : it is not yet supported to write to hive partitions with datatype Dictionary(UInt16, Utf8) #7891
• writing to partitioned table uses the wrong column as partition key #7892
• Support writing hive style partitioned files in COPY command #8493
• Parallel NDSON file reading #8502
• Parallel Arrow file format reading #8503
• Support writing to Arrow files #8504
• Remove External Table Backwards Compatibility Options #8547
• Remove SINGLE_FILE_OUTPUT option from COPY statement #8621
• Support Configuring arrow_ipc::FileWriter via ArrowWriterOptions #8635
• Make/Remove FileType enum and replace with a trait #8657
• Support writing hive style partitioned files in DataFrame::write command #9237

[alamb]

FYI @devinjdangelo I updated this ticket with various issues related to the write code you are working on

[devinjdangelo]

@alamb I opened #7298 to track improving statement level options/overrides relevant to this epic.

[alamb]

@alamb I opened #7298 to track improving statement level options/overrides relevant to this epic.

Thanks @devinjdangelo -- I added it to the list

[devinjdangelo]

An additional issue we should cut and add to this epic is allowing inserts to a sorted ListingTable. In the case of appending new files to a directory, I think it is as simple as having FileSinkExec require its input be sorted.

It can't really be supported efficiently for Append to existing file since it would require reading the existing file, sorting with the new data and rewriting the whole file. For this case, you could use insert overwrite instead if you really want to do this (which is another thing which we could cut a ticket to add support for).

Alternatively, we could have a check to see if 1) the table is sorted and 2) the input to FileSinkExec is sorted. If 1) is true but 2) is not, we would need to update the metadata about the table to indicate for subsequent queries it is no longer guaranteed to be sorted.

[alamb]

Filed #7354 to track

[devinjdangelo]

@alamb I opened #7442 to track adding support for parquet column level settings via SQL options.

[alamb]

@alamb I opened #7442 to track adding support for parquet column level settings via SQL options.

Thanks @devinjdangelo -- I added it to the list on this ticket

[devinjdangelo]

I added #7679 to track adding avro support for writes @alamb

[devinjdangelo]

@alamb I made some progress on inserts to sorted tables #7354

This also got me thinking about inserts to partitioned tables, so I opened issue to track:
#7744

Lastly, I've been thinking we may want to deprecate and eventually remove the SessionContext write methods or alternatively hook them into FileSinkExec similarly to how the data frame write methods are set up.
https://github.com/apache/arrow-datafusion/blob/0408c2b1596417ba55a636fa3c8a601ffbdb0e60/datafusion/core/src/execution/context.rs#L1271-L1278

[alamb]

This also got me thinking about inserts to partitioned tables, so I opened issue to track: #7744

Thank you -- I added #7744 to the list on this ticket

Lastly, I've been thinking we may want to deprecate and eventually remove the SessionContext write methods or alternatively hook them into FileSinkExec similarly to how the data frame write methods are set up.

That sounds like a reasonable idea to me. One challenge might be that the SessionContext::write_csv take an ExecutionPlan where the apis on DataFrame require a LogicalPlan.

Hooking them into FileSinkExec sounds like a great idea to consolidate the code. Shall I file a ticket for it?

[devinjdangelo]

@alamb I wrote up an issue describing what we discussed on #7743 regarding empty files being written out and some potential solutions.

#7767