Spark Expression Support#

This page is the complete reference for how Apache Comet handles each Spark built-in expression. Comet accelerates expressions either with a native (Rust) implementation or by dispatching to a Spark-compatible codegen path. When an expression is not supported, Comet transparently falls back to Spark for that part of the plan; results are unaffected.

Expressions marked ✅ Supported are enabled by default and produce Spark-compatible results.

Some ✅ Supported expressions have specific incompatible cases that are not run by default. Those cases must be opted into per expression with spark.comet.expression.EXPRNAME.allowIncompatible=true (where EXPRNAME is the Spark expression class name, for example Cast). There is no global opt-in. By default such a case either falls back to Spark (for example cast) or, when the expression has a Spark-compatible codegen-dispatch implementation, runs through that instead (for example the regex and JSON families). See Native and codegen-dispatch implementations for how Comet chooses.

Most expressions can also be disabled with spark.comet.expression.EXPRNAME.enabled=false, where EXPRNAME is the Spark expression class name (for example Length or StartsWith). See the Comet Configuration Guide for the full list.

Status legend#

Status	Meaning
✅ Supported	Comet produces Spark-compatible results by default. Some inputs or forms may fall back to Spark, and any incompatible behavior is opt-in (off by default).
🔜 Planned	Intended; tracked by an open issue or pull request.

Not currently planned#

Comet focuses acceleration on mainstream relational, string, datetime, math, and collection expressions. The following function families are not currently planned for native acceleration (they are not on the 1.0 roadmap): specialized functionality with narrow real-world analytics use and high implementation cost. They fall back to Spark and may be reconsidered based on demand:

Probabilistic sketches and approximate top-k (kll_sketch_*, hll_*, theta_*, count_min_sketch, bitmap_*, approx_top_k*): specialized data structures with exact-correctness traps.
Geospatial (st_*): brand-new Spark 4.1 functionality, specialized.
Avro / Protobuf codecs (from_avro, to_avro, from_protobuf, to_protobuf, schema_of_avro): format conversion belongs at the IO layer, not expression evaluation.
JVM reflection (java_method, reflect): niche, and they invoke arbitrary JVM methods (a security concern).
UTF-8 validation (is_valid_utf8, make_valid_utf8, validate_utf8, try_validate_utf8): niche Spark 4.x string-validation helpers.
Miscellaneous niche (histogram_numeric, version, sentences, quote): low-value or specialized functions with little benefit from native acceleration.

The file-metadata functions input_file_name, input_file_block_start, and input_file_block_length depend on scan-internal per-row file information rather than the expression layer; their support status is covered in the scan compatibility guide.

Note that median and mode are planned: they are mainstream exact aggregates. approx_count_distinct is supported because Comet ports Spark’s HyperLogLogPlusPlus exactly, so its result is bit-identical to Spark.

The tables below list every Spark built-in expression with its current status.

agg_funcs#

Function	Status	Notes
`any`	✅
`any_value`	✅
`approx_count_distinct`	✅
`approx_percentile`	✅	Byte, short, int, long, float, and double input; other input types fall back to Spark
`array_agg`	🔜	Array aggregate (related to `collect_list`, #2524)
`avg`	✅	Interval types fall back
`bit_and`	✅
`bit_or`	✅
`bit_xor`	✅
`bool_and`	✅
`bool_or`	✅
`collect_list`	🔜	#2524
`collect_set`	✅
`corr`	✅
`count`	✅
`count_if`	✅
`covar_pop`	✅
`covar_samp`	✅
`every`	✅
`first`	✅
`first_value`	✅
`grouping`	✅	Grouping indicator for ROLLUP/CUBE/GROUPING SETS
`grouping_id`	✅	Grouping indicator for ROLLUP/CUBE/GROUPING SETS
`kurtosis`	🔜	tracking #4098
`last`	✅
`last_value`	✅
`listagg`	🔜	String aggregation
`max`	✅
`max_by`	🔜	#3841
`mean`	✅
`median`	✅	Rewrites to `percentile(col, 0.5)` and runs natively for supported percentile inputs
`min`	✅
`min_by`	🔜	#3841
`mode`	🔜	#3970
`percentile`	✅	Single literal percentage on numeric input runs natively; array of percentages and a frequency argument fall back to Spark
`percentile_cont`	✅	Spark 4.0+ `WITHIN GROUP (ORDER BY ...)`; ascending only runs natively, `DESC` falls back to Spark
`percentile_disc`	🔜	Percentile aggregate
`regr_avgx`	✅	Native: Spark rewrites to `Average` (tests in #4551)
`regr_avgy`	✅	Native: Spark rewrites to `Average` (tests in #4551)
`regr_count`	✅	Native: Spark rewrites to `Count` (tests in #4551)
`regr_intercept`	🔜	Falls back; can reuse `covar_pop`/`var_pop` accumulators (#4552)
`regr_r2`	🔜	Falls back; can reuse the `corr` accumulator (#4552)
`regr_slope`	🔜	Falls back; can reuse `covar_pop`/`var_pop` accumulators (#4552)
`regr_sxx`	🔜	Falls back; can reuse `var_pop` accumulator (#4552)
`regr_sxy`	🔜	Falls back; can reuse `covar_pop` accumulator (#4552)
`regr_syy`	🔜	Falls back; can reuse `var_pop` accumulator (#4552)
`skewness`	🔜	tracking #4098
`some`	✅
`std`	✅
`stddev`	✅
`stddev_pop`	✅
`stddev_samp`	✅
`string_agg`	🔜	String aggregation (alias of `listagg`)
`sum`	✅
`try_avg`	✅	Interval types fall back
`try_sum`	✅
`var_pop`	✅
`var_samp`	✅
`variance`	✅

array_funcs#

Function	Status	Notes
`array`	✅
`array_append`	✅
`array_compact`	✅
`array_contains`	✅	NaN/signed-zero handling may differ (details)
`array_distinct`	✅	NaN/signed-zero handling may differ (details)
`array_except`	✅	Routes through the JVM codegen dispatcher by default; the incompatible native path is opt-in via allowIncompatible (details)
`array_insert`	✅
`array_intersect`	✅	Routes through the JVM codegen dispatcher by default; the incompatible native path is opt-in via allowIncompatible (details)
`array_join`	✅	Routes through the JVM codegen dispatcher by default; the incompatible native path is opt-in via allowIncompatible (details)
`array_max`	✅	NaN ordering may differ (details)
`array_min`	✅	NaN ordering may differ (details)
`array_position`	✅	Binary/struct/map/null elements fall back
`array_prepend`	✅
`array_remove`	✅
`array_repeat`	✅
`array_union`	✅	NaN/signed-zero handling may differ (details)
`arrays_overlap`	✅
`arrays_zip`	✅
`element_at`	✅
`flatten`	✅	Binary/struct/map elements fall back
`get`	✅
`sequence`	✅
`shuffle`	✅	Binary/struct/map elements fall back
`slice`	✅	Native (#4149)
`sort_array`	✅	Nested struct/null arrays fall back

bitwise_funcs#

Function	Status	Notes
`&`	✅
`<<`	✅
`>>`	✅
`>>>`	✅	Operator alias for `shiftrightunsigned` (Spark 4.0+)
`^`	✅
`bit_count`	✅
`bit_get`	✅
`getbit`	✅
`shiftright`	✅
`shiftrightunsigned`	✅
`\|`	✅
`~`	✅

collection_funcs#

Function	Status	Notes
`array_size`	✅
`cardinality`	✅
`concat`	✅	Binary/array children fall back
`reverse`	✅	Binary-element arrays fall back (Incompatible) (details)
`size`	✅

conditional_funcs#

Function	Status	Notes
`coalesce`	✅
`if`	✅
`ifnull`	✅
`nanvl`	✅
`nullif`	✅
`nullifzero`	✅	Lowers to `if`/`=` (Spark 4.0+)
`nvl`	✅
`nvl2`	✅
`when`	✅
`zeroifnull`	✅	Lowers to `coalesce` (Spark 4.0+)

conversion_funcs#

The type-name conversion functions (bigint, binary, boolean, date, decimal, double, float, int, smallint, string, timestamp, tinyint) are SQL aliases for CAST(... AS <type>) and share the support and caveats of cast.

Function	Status	Notes
`cast`	✅	Some casts fall back; float-to-decimal is opt-in (details)

csv_funcs#

Function	Status	Notes
`from_csv`	✅
`schema_of_csv`	✅
`to_csv`	✅

datetime_funcs#

Function	Status	Notes
`add_months`	✅
`convert_timezone`	✅	Routes through the JVM codegen dispatcher by default (handles all timezone forms); the native path is opt-in via allowIncompatible (details)
`curdate`	✅	Constant-folded to a literal (alias of `current_date`)
`current_date`	✅	Constant-folded to a literal before Comet sees the plan
`current_time`	🔜	Blocked on Spark 4.1 TIME type support (#4288)
`current_timestamp`	✅	Constant-folded to a literal before Comet sees the plan
`current_timezone`	✅
`date_add`	✅
`date_diff`	✅
`date_format`	✅
`date_from_unix_date`	✅
`date_part`	✅
`date_sub`	✅
`date_trunc`	✅
`dateadd`	✅
`datediff`	✅
`datepart`	✅
`day`	✅
`dayname`	✅	Abbreviated day name (Spark 4.0+)
`dayofmonth`	✅
`dayofweek`	✅
`dayofyear`	✅
`extract`	✅
`from_unixtime`	✅
`from_utc_timestamp`	✅	Routes through the JVM codegen dispatcher by default (handles all timezone forms); the native path is opt-in via allowIncompatible (details)
`hour`	✅
`last_day`	✅
`localtimestamp`	✅
`make_date`	✅
`make_dt_interval`	✅
`make_interval`	🔜	Produces legacy CalendarInterval; tracked by #4540
`make_time`	🔜	Spark 4.1 TIME type; tracked by #4288
`make_timestamp`	✅
`make_timestamp_ltz`	✅	2-arg TIME form falls back
`make_timestamp_ntz`	✅	2-arg TIME form falls back
`make_ym_interval`	✅
`minute`	✅
`month`	✅
`monthname`	✅	Abbreviated month name (Spark 4.0+)
`months_between`	✅
`next_day`	✅
`now`	✅	Constant-folded to a literal (alias of `current_timestamp`)
`quarter`	✅
`second`	✅
`session_window`	🔜	Batch session-window grouping falls back (`UpdatingSessionsExec` is not yet native); tracked by #4785
`time_diff`	🔜	Spark 4.1 TIME type; tracked by #4288
`time_trunc`	🔜	Spark 4.1 TIME type; tracked by #4288
`timestamp_micros`	✅
`timestamp_millis`	✅
`timestamp_seconds`	✅
`to_date`	✅	Rewrites to `Cast` (or `Cast(GetTimestamp)` with a format) before Comet sees the plan
`to_time`	🔜	Spark 4.1 TIME type; tracked by #4288
`to_timestamp`	✅	Rewrites to `Cast` (or `GetTimestamp` with a format) before Comet sees the plan
`to_timestamp_ltz`	✅	Rewrites to `to_timestamp` (`TimestampType`)
`to_timestamp_ntz`	✅	Rewrites to `to_timestamp` (`TimestampNTZType`)
`to_unix_timestamp`	✅
`to_utc_timestamp`	✅	Routes through the JVM codegen dispatcher by default (handles all timezone forms); the native path is opt-in via allowIncompatible (details)
`trunc`	✅
`try_make_interval`	🔜	Produces legacy CalendarInterval; tracked by #4540
`try_make_timestamp`	✅
`try_to_date`	✅	Rewrites to `Cast`/`GetTimestamp` before Comet sees the plan; same support as `to_date`
`try_to_time`	🔜	Spark 4.1 TIME type; tracked by #4288
`try_to_timestamp`	✅	Rewrites to `Cast`/`GetTimestamp` before Comet sees the plan; same support as `to_timestamp`
`unix_date`	✅
`unix_micros`	✅
`unix_millis`	✅
`unix_seconds`	✅
`unix_timestamp`	✅
`weekday`	✅
`weekofyear`	✅
`window`	✅	Batch tumbling and sliding time-window grouping runs natively
`window_time`	✅	Batch time-window grouping runs natively
`year`	✅

generator_funcs#

explode and posexplode are supported via CometExplodeExec (operator-level, not expression-level). The outer variants are wired but marked Incompatible; they require spark.comet.exec.explode.enabled=true and allowIncompatible.

Function	Status	Notes
`explode`	✅	via `CometExplodeExec`
`explode_outer`	✅	outer=true falls back (Incompatible) (audit)
`inline`	🔜	Operator-level generator (like `explode`)
`inline_outer`	🔜	Operator-level generator (like `explode`)
`posexplode`	✅	via `CometExplodeExec`
`posexplode_outer`	✅	outer=true falls back (Incompatible) (audit)
`stack`	🔜	Operator-level generator

hash_funcs#

Function	Status	Notes
`crc32`	✅
`hash`	✅
`md5`	✅
`sha`	✅
`sha1`	✅
`sha2`	✅
`xxhash64`	✅

json_funcs#

Function	Status	Notes
`from_json`	✅	Falls back by default; opt-in via allowIncompatible (audit)
`get_json_object`	✅	Some inputs need allowIncompatible (audit)
`json_array_length`	✅	Single-quoted/trailing JSON needs allowIncompatible (audit)
`json_object_keys`	✅
`json_tuple`	🔜	#3160
`schema_of_json`	✅
`to_json`	✅	Options and map/array inputs fall back (audit)

lambda_funcs#

Function	Status	Notes
`aggregate`	✅
`array_sort`	✅
`exists`	✅
`filter`	✅	General lambda routed through the JVM codegen dispatcher; the `array_compact` form runs natively
`forall`	✅
`map_filter`	✅
`map_zip_with`	✅
`reduce`	✅
`transform`	✅
`transform_keys`	✅
`transform_values`	✅
`zip_with`	✅

map_funcs#

Function	Status	Notes
`element_at`	✅
`map`	✅	Routed through the JVM codegen dispatcher
`map_concat`	✅
`map_contains_key`	✅
`map_entries`	✅
`map_from_arrays`	✅
`map_from_entries`	✅	BinaryType key/value falls back (Incompatible) (details)
`map_keys`	✅
`map_values`	✅
`str_to_map`	✅
`try_element_at`	✅	Lowers to `element_at`

math_funcs#

Function	Status	Notes
`%`	✅
`*`	✅	Interval multiplication falls back
`+`	✅
`-`	✅
`/`	✅
`abs`	✅	Interval types fall back
`acos`	✅
`acosh`	✅
`asin`	✅
`asinh`	✅
`atan`	✅
`atan2`	✅
`atanh`	✅
`bin`	✅
`bround`	✅
`cbrt`	✅
`ceil`	✅	Two-arg form falls back
`ceiling`	✅
`conv`	✅
`cos`	✅
`cosh`	✅
`cot`	✅
`csc`	✅
`degrees`	✅
`div`	✅
`e`	✅	Folds to a literal (like `pi`)
`exp`	✅
`expm1`	✅
`factorial`	✅
`floor`	✅	Two-arg form falls back
`greatest`	✅
`hex`	✅
`hypot`	✅
`least`	✅
`ln`	✅
`log`	✅
`log10`	✅
`log1p`	✅
`log2`	✅
`mod`	✅
`negative`	✅
`pi`	✅
`pmod`	✅
`positive`	✅
`pow`	✅
`power`	✅
`radians`	✅
`rand`	✅
`randn`	✅
`random`	✅	Alias for `rand` (Spark 4.0+); seed must be a literal
`randstr`	🔜	Random string (Spark 4.0+)
`rint`	✅
`round`	✅	Float/double inputs fall back
`sec`	✅
`shiftleft`	✅
`sign`	✅
`signum`	✅
`sin`	✅
`sinh`	✅
`sqrt`	✅
`tan`	✅
`tanh`	✅
`try_add`	✅	Datetime/interval form falls back
`try_divide`	✅
`try_mod`	✅
`try_multiply`	✅
`try_subtract`	✅
`unhex`	✅
`uniform`	✅	Constant-folded; literal arguments only (Spark 4.0+)
`width_bucket`	✅

misc_funcs#

Function	Status	Notes
`aes_decrypt`	✅	Routed through the JVM codegen dispatcher
`aes_encrypt`	✅	Routed through the JVM codegen dispatcher; nondeterministic IV by default
`assert_true`	🔜	Lowers to `RaiseError`, which falls back
`current_catalog`	✅	Resolved to a literal by the analyzer (`ReplaceCurrentLike`)
`current_database`	✅	Resolved to a literal by the analyzer (`ReplaceCurrentLike`)
`current_schema`	✅	Alias of `current_database`; resolved to a literal by the analyzer
`current_user`	✅	Resolved to a literal by the analyzer; same as `user`
`equal_null`	✅	Lowers to `<=>` (`EqualNullSafe`)
`is_variant_null`	🔜	tracking #4098
`monotonically_increasing_id`	✅
`parse_json`	🔜	tracking #4098
`raise_error`	🔜	Raises a runtime error
`rand`	✅	Seed must be a literal
`randn`	✅	Seed must be a literal
`schema_of_variant`	🔜	tracking #4098
`schema_of_variant_agg`	🔜	tracking #4098
`session_user`	✅	Alias of `current_user`; resolved to a literal by the analyzer
`spark_partition_id`	✅
`to_variant_object`	🔜	tracking #4098
`try_aes_decrypt`	✅	Routed through the JVM codegen dispatcher
`try_parse_json`	🔜	tracking #4098
`try_variant_get`	🔜	tracking #4098
`typeof`	✅	Foldable; resolved to a literal before Comet sees the plan
`user`	✅	Resolved to a literal by the Spark analyzer before reaching Comet
`uuid`	🔜	Nondeterministic random UUID
`variant_get`	🔜	tracking #4098

predicate_funcs#

Function	Status	Notes
`!`	✅
`<`	✅
`<=`	✅
`<=>`	✅
`=`	✅
`==`	✅
`>`	✅
`>=`	✅
`and`	✅
`between`	✅
`ilike`	✅
`in`	✅
`isnan`	✅
`isnotnull`	✅
`isnull`	✅
`like`	✅
`not`	✅
`or`	✅
`regexp`	✅	Falls back by default; opt-in via allowIncompatible (details)
`regexp_like`	✅	Falls back by default; opt-in via allowIncompatible (details)
`rlike`	✅	Falls back by default; opt-in via allowIncompatible (details)

string_funcs#

Function	Status	Notes
`ascii`	✅
`base64`	✅
`bit_length`	✅
`btrim`	✅
`char`	✅
`char_length`	✅
`character_length`	✅
`chr`	✅
`collate`	🔜	Spark collation (umbrella #2190)
`collation`	✅	Constant-folded to a literal (Spark 4.0+)
`concat_ws`	✅
`contains`	✅
`decode`	✅
`elt`	✅
`encode`	🔜	Lowers to `StaticInvoke(encode)` (not allowlisted); falls back
`endswith`	✅
`find_in_set`	✅
`format_number`	✅
`format_string`	✅
`initcap`	✅
`instr`	✅
`lcase`	✅
`left`	✅
`len`	✅
`length`	✅
`levenshtein`	✅
`locate`	✅
`lower`	✅
`lpad`	✅
`ltrim`	✅
`luhn_check`	✅	Native via `StaticInvoke` (tests: luhn_check.sql)
`mask`	✅	Routed through the JVM codegen dispatcher
`octet_length`	✅
`overlay`	✅
`position`	✅
`printf`	✅
`regexp_count`	✅	Runs natively (rewrites to `size(regexp_extract_all(...))`)
`regexp_extract`	✅
`regexp_extract_all`	✅
`regexp_instr`	✅	Routed through the JVM codegen dispatcher
`regexp_replace`	✅
`regexp_substr`	✅	Runs natively (rewrites to `nullif(regexp_extract(...), '')`)
`repeat`	✅
`replace`	✅
`right`	✅
`rpad`	✅
`rtrim`	✅
`soundex`	✅
`space`	✅
`split`	✅
`split_part`	✅	Spark 4.0+
`startswith`	✅
`substr`	✅
`substring`	✅
`substring_index`	✅
`to_binary`	✅	Hex form accelerated; other formats fall back
`to_char`	✅
`to_number`	✅
`to_varchar`	✅
`translate`	✅	Falls back by default; opt-in via allowIncompatible (#4463)
`trim`	✅
`try_to_binary`	✅	Runs natively (rewrites to `try_eval(to_binary(...))`)
`try_to_number`	✅	Routed through the JVM codegen dispatcher
`ucase`	✅
`unbase64`	✅
`upper`	✅

struct_funcs#

Function	Status	Notes
`named_struct`	✅	Duplicate field names fall back
`struct`	✅

url_funcs#

Function	Status	Notes
`parse_url`	✅
`try_url_decode`	✅
`url_decode`	✅
`url_encode`	✅

window_funcs#

Window functions run via CometWindowExec, which is enabled by default. Aggregate window functions (count, min, max, sum, avg, first_value, last_value), ranking functions (row_number, rank, dense_rank, percent_rank, cume_dist, ntile), and value-shift functions (lag, lead, nth_value) are all wired in the window serde and execute natively. Statistical aggregates such as stddev, var_pop, corr, and covar_pop run natively as plain aggregations but fall back to Spark when used as window functions. A handful of frame shapes also fall back. See window function compatibility for the full list of supported functions, frames, and fallback cases.

Function	Status	Notes
`cume_dist`	✅	via `CometWindowExec`
`dense_rank`	✅	via `CometWindowExec`
`lag`	✅	via `CometWindowExec`; non-literal default falls back (#4268)
`lead`	✅	via `CometWindowExec`; non-literal default falls back (#4268)
`nth_value`	✅	via `CometWindowExec`
`ntile`	✅	via `CometWindowExec`
`percent_rank`	✅	via `CometWindowExec`
`rank`	✅	via `CometWindowExec`
`row_number`	✅	via `CometWindowExec`

xml_funcs#

Function	Status	Notes
`from_xml`	✅	Spark 4.0+
`schema_of_xml`	✅	Spark 4.0+
`to_xml`	✅	Spark 4.0+
`xpath`	✅
`xpath_boolean`	✅
`xpath_double`	✅
`xpath_float`	✅
`xpath_int`	✅
`xpath_long`	✅
`xpath_number`	✅	Alias of `xpath_double`
`xpath_short`	✅
`xpath_string`	✅

Beyond SQL functions#

Comet also accelerates a number of Catalyst expressions that have no Spark SQL function name and therefore do not appear in the tables above. These arise from the DataFrame API, from SQL syntax other than function calls, or from the query optimizer. They include:

Operator and optimizer-injected expressions: runtime bloom-filter join probes (BloomFilterMightContain, BloomFilterAggregate), optimized IN sets (InSet), scalar subqueries (ScalarSubquery), and floating-point normalization (KnownFloatingPointNormalized).
Accessor expressions (subscript and field access, not functions): struct field access (col.field), array element access (arr[i]), and map value access (map[key]).
Internal decimal arithmetic: CheckOverflow, MakeDecimal, and UnscaledValue, which the analyzer inserts around decimal operations.
User-defined functions: Scala UDFs registered through the DataFrame or SQL API.
Structural expressions: aliases, attribute references, literals, sort orders, and CASE WHEN.

This list is illustrative, not exhaustive: the per-function tables are not the complete set of expressions Comet can accelerate.