feat: close upstream coverage gaps for DataFusion 55.1.0 - #1763

Merged

timsaucer merged 59 commits into

apache:mainfrom

timsaucer:feat/upstream-coverage-gaps

Oct 4, 2026

Commits on Sep 25, 2026

feat: expose any_value, array math, array_first, and missing Spark functions

timsaucer
and
claude
committed
5a0e79d
feat: add rand, substring_index, file metadata functions, and distinct kwargs

timsaucer
and
claude
committed
8027fcb
feat(spark): add pyspark aliases and optional-length substr

timsaucer
and
claude
committed
97e6031
feat: accept optional arguments upstream supports on trim, array_to_string, substr

timsaucer
and
claude
committed
e8c2192
feat: add DataFrame.fill_nan

timsaucer
and
claude
committed
2747573
feat: add null_treatment to lead and lag

timsaucer
and
claude
committed
24764bd
feat: add show_statistics, analyze_level, and analyze_categories to explain

timsaucer
and
claude
committed
b035c97
fix: keep existing options when chaining aggregate and window builders

timsaucer
and
claude
committed
d2ac890
feat: accept a bare PyCapsule in ScalarUDF and WindowUDF

timsaucer
and
claude
committed
97e5905
chore: export TableProviderFactory from datafusion.catalog

timsaucer
and
claude
committed
d53a255
docs(skill): record DataFrame.to_string as not needing exposure

timsaucer
and
claude
committed
5d43576
fix: resolve the _PyCapsule type alias so capsule overloads type-check

timsaucer
and
claude
committed
f6d0478
fix: keep an explicit window frame that equals the default when chaining

timsaucer
and
claude
committed
6345b17
fix: leave lead and lag null_treatment unset by default

timsaucer
and
claude
committed
73b052d
fix: keep existing window function options in over()

timsaucer
and
claude
committed
8173554
feat: accept native ints and a single argument in range and gen_series

timsaucer
and
claude
committed
96f990b

Commits on Sep 28, 2026

fix: decide window frame re-derivation from the expression alone

timsaucer
and
claude
committed
d731279
fix: treat the frame an empty order_by derives as a default when chaining

timsaucer
and
claude
committed
b4cad2d
fix: raise when a builder option does not apply to the function kind

timsaucer
and
claude
committed
92ace16
docs: note that over() drops options set on an aggregate

timsaucer
and
claude
committed
8e21c8a
docs: document that chaining now keeps options already set

timsaucer
and
claude
committed
6b52225
fix: reject a non-bool distinct in string_agg

timsaucer
and
claude
committed
0ac494e
fix: reject show_statistics combined with analyze in explain

timsaucer
and
claude
committed
607dde4
fix: treat a bare str as a column name in spark.printf

timsaucer
and
claude
committed
4933fa2
fix: name the expected and found capsule when importing a UDF

timsaucer
and
claude
committed
23ed533
fix: treat an empty fill_null or fill_nan subset as no columns

timsaucer
and
claude
committed
8454a09
fix: keep the sort direction in percentile_cont and friends

timsaucer
and
claude
committed
1b01845
fix: allow the keyword-only form of the udf, udaf, and udwf decorators

timsaucer
and
claude
committed
f8d7b70
docs: fix the broken guide links in aggregate docstrings

timsaucer
and
claude
committed
7259819
test: check that bit_and and bit_or keep distinct in the expression

timsaucer
and
claude
committed
d779502
docs(skill): show substr's length argument as valid

timsaucer
and
claude
committed
db5b7cc

Commits on Sep 30, 2026

fix: keep aggregate options in over() and reject an aggregate order_by

timsaucer
and
claude
committed
2fb7cce
fix: treat the sort key added when decoding a RANGE frame as no order_by

timsaucer
and
claude
committed
99f506d
fix: check the function kind on every builder call, not just the first

timsaucer
and
claude
committed
7dd2658
fix: accept numpy and pandas column lists as fill_null and fill_nan subsets

timsaucer
and
claude
committed
e811758
fix: reject explain options that the chosen plan would ignore

timsaucer
and
claude
committed
1af877f
fix: take a lone range or gen_series argument only by position

timsaucer
and
claude
committed
473856e
fix: reject DISTINCT in Python aggregate UDFs instead of ignoring it

timsaucer
and
claude
committed
74b6434
docs: say a chained order_by sets only the percentile's sort direction

timsaucer
and
claude
committed
daac7f2
docs: correct what distinct does in count_star and the weighted percentile

timsaucer
and
claude
committed
a4fe393
fix: accept numpy.bool_ as string_agg's distinct

timsaucer
and
claude
committed
47359df
fix: accept the callable by keyword in udf, udaf, udwf, and udtf

timsaucer
and
claude
committed
1475123
docs: note that fill_null and fill_nan fail on uppercase or dotted names

timsaucer
and
claude
committed
354db9a
docs: say mean's filter now works only when passed by keyword

timsaucer
and
claude
committed
9f2b326
fix: resolve CapsuleType in context and extensions, and at runtime

timsaucer
and
claude
committed
6808f95
docs: show how to move an aggregate's order_by into the Window

timsaucer
and
claude
committed
4d66257
fix: import CapsuleType from types on Python 3.13 and newer

timsaucer
and
claude
committed
ad4a447
fix: require typing-extensions 4.12 or newer below Python 3.13

timsaucer
and
claude
committed
cada104
fix: reject an explicit None stop in range and gen_series

timsaucer
and
claude
committed
6ae9e6e
fix: treat an empty window order_by as no order_by

timsaucer
and
claude
committed
61cb60f
fix: name the function when a positional filter lands in distinct

timsaucer
and
claude
committed
ec34c78
refactor: make stop the required argument of range and gen_series

timsaucer
and
claude
committed
86b4356
refactor: build capsule-backed UDFs through _from_internal

timsaucer
and
claude
committed
349c17b
refactor: derive builder kind and options in one match

timsaucer
and
claude
committed
487ab3c
docs: drop the claim that every ExprFuncBuilder method raises

timsaucer
and
claude
committed
020de42

Commits on Oct 1, 2026

test: drop low-value cases and fold duplicates added on this branch

timsaucer
and
claude
committed
8fe4188
fix: raise when chaining onto a window function that has window options

timsaucer
and
claude
committed
5d79982
docs: correct how to set null treatment on a window and list array_first

timsaucer
and
claude
committed
74315c2
docs: clarify which DISTINCT queries reach a Python UDAF without the flag

timsaucer
and
claude
committed
a0008f3