Set the default value of datafusion.execution.collect_statistics to true

[AdamGS]

Which issue does this PR close?

• Closes #Default to collecting statistics when creating LIstingTables #16158 and Datafusion 48 Clickbench Q6 and Q0 regression #16444.

Rationale for this change

What changes are included in this PR?

Are these changes tested?

Are there any user-facing changes?

[jonathanc-n]

I think we should run some benchmarks here to see if the regression is gone

[AdamGS]

Definitely, I'll run our benchmarks once I get all tests passing here.

[AdamGS]

I'm getting a lot of sqllogictest failures, is there a reason to think there something weird going on? I was somewhat open to the idea its all fine until I ran into the last test in union_by_name.slt suddenly passing.

I'll push all of them soon so others can also take a look.

[alamb]

Thank you @AdamGS !

I think the reason that so many of the plans changed is that there is rule that skips repartitioning input files if they are past some threshold (repartition_file_min_size) and now that the statistics are enabled by default this code is triggered

[alamb]

Thank you for this contribution @AdamGS -- this looks great to me and I think the plan changes are also good / reasonable.

I think we should also add a note to the upgrade guide about this change: https://github.com/apache/datafusion/blob/main/docs/source/library-user-guide/upgrading.md

FYI @davisp

[alamb]

Note to other reviewers, the negative is still tested below:

When the setting is false, the statistics are NOT gathered

[AdamGS]

Added a short upgrade note

[AdamGS]

Got a similar test failure to #16448 (issue filed in #16452). I have to conclude its personal at this point, I'll try and find some time to dig into it.

[AdamGS]

Our benchmarks show this change fixes the performance regression we saw - vortex-data/vortex#3567

[Copilot]

Pull Request Overview

This PR changes the default of datafusion.execution.collect_statistics from false to true, updating documentation, configuration defaults, upgrade notes, and test expectations to match the new behavior.

• Update user‐guide and config docs to reflect that statistics are now gathered by default.
• Amend upgrade guide and example override to the new default.
• Adjust code comments, configuration struct, and tests (unit and SQL logic tests) for default‐true behavior.

Reviewed Changes

Copilot reviewed 12 out of 12 changed files in this pull request and generated no comments.

Show a summary per file

File	Description
docs/source/user-guide/sql/ddl.md	Note updated to describe reading files by default
docs/source/user-guide/configs.md	Default value for collect_statistics set to true
docs/source/library-user-guide/upgrading.md	Upgrade notes and example override updated
datafusion/sqllogictest/test_files/repartition.slt	Physical plan ordering updated
datafusion/sqllogictest/test_files/parquet_statistics.slt	Expected stats presence in plan updated
datafusion/sqllogictest/test_files/parquet_filter_pushdown.slt	Removed redundant nodes, fixed spelling
datafusion/sqllogictest/test_files/limit.slt	Stripped byte-range suffixes from file paths
datafusion/sqllogictest/test_files/information_schema.slt	Schema output default toggled to true
datafusion/sqllogictest/test_files/explain_tree.slt	Regenerated explain output formatting
datafusion/core/tests/parquet/row_group_pruning.rs	Adjusted expected pruned_by_stats counts
datafusion/core/src/execution/context/parquet.rs	Doc comments and test assertions updated
datafusion/common/src/config.rs	Config default and doc comment updated

Comments suppressed due to low confidence (2)

docs/source/library-user-guide/upgrading.md:30

• [nitpick] The sentence is incomplete. Consider rephrasing to clarify what is 'previous' (e.g., "restores the previous default behavior of ListingTable").

This change also restores the default behavior of `ListingTable` to its previous. If you use it directly

docs/source/library-user-guide/upgrading.md:36

• Verify that the builder method name matches the actual API. If it's named with_collect_statistics rather than with_collect_stat, update this example accordingly.

    .with_collect_stat(false)

[blaginin]

Fuzzy failed again, presumably due to #16452 - rerun resolved it