Repository navigation
Spark, Databricks: port upstream interval PRs #2614 and #2616 - #19
Merged
moshap-firebolt merged 9 commits intoOct 22, 2026
Merged
Conversation
This reverts commit bae1cb2.
)" This reverts commit ee51174.
Spark's multi-units interval syntax lets the string carry its own units, as in INTERVAL '1 YEAR 2 DAYS 3 HOURS', so no unit follows the literal. Databricks Runtime parses with the same grammar. A unit after a literal is still read; the expression form this gives up (INTERVAL 1 + 1 DAY) is not Spark syntax.
Address review: turning require_interval_qualifier off accepted INTERVAL 1 and read SELECT interval FROM t as SELECT INTERVAL FROM AS t. Keep it on and add supports_interval_string_without_qualifier: the value must be a literal, and only an unsigned string may omit the qualifier. INTERVAL is not reserved for these dialects, so SELECT interval FROM t reads a column.
…ter interval is arithmetic Address review: a double-quoted string is a Spark string literal, so it may be an interval value and may omit the qualifier (INTERVAL "2 months", INTERVAL "1" DAY). Only a bare number or a string after INTERVAL keeps the interval error; a signed number falls back to the identifier, so interval - 1 and INTERVAL -1 are arithmetic on the column, as in Spark. A string window frame bound stays a string literal in these dialects (RANGE BETWEEN '1 day' PRECEDING), as in Spark, rather than being read as an interval without a qualifier.
Address review: Spark accepts INTERVAL '1' DAY TO as an alias, so assert the error on TO x instead. parse_interval_number_without_unit_error and parse_interval_expression_value repeat test_interval_numeric_requires_unit and parse_interval_require_qualifier. The guard comment now covers strings as well as numbers.
Spark's multi-units interval syntax lists several value / unit pairs, INTERVAL 10 YEAR 20 MONTH, documented as the same as the string form INTERVAL '10 YEAR 20 MONTH'. A new supports_interval_multi_units hook, enabled for SparkSqlDialect and DatabricksDialect, reads the extra pairs into that string form, so the AST is unchanged.
A double-quoted string is a Spark string literal, so it may be any value of a multi-units interval, e.g. INTERVAL "1" DAY "2" HOUR.
moshap-firebolt
force-pushed
the
moshap/port-upstream-intervals
branch
from
October 20, 2026 01:02
c0c01aa to
a70bb84
Compare
…ue_arithmetic clippy's assertions_on_result_states rejects assert!(... .is_err()). Compare the message instead, as test_interval_numeric_requires_unit does.
This file contains hidden or bidirectional Unicode text that may be interpreted or compiled differently than what appears below. To review, open the file in an editor that reveals hidden Unicode characters.
Learn more about bidirectional Unicode characters
Sign up for free
to join this conversation on GitHub.
Already have an account?
Sign in to comment
1 participant
Add this suggestion to a batch that can be applied as a single commit.This suggestion is invalid because no changes were made to the code.Suggestions cannot be applied while the pull request is closed.Suggestions cannot be applied while viewing a subset of changes.Only one suggestion per line can be applied in a batch.Add this suggestion to a batch that can be applied as a single commit.Applying suggestions on deleted lines is not supported.You must change the existing code in this line in order to create a valid suggestion.Outdated suggestions cannot be applied.This suggestion has been applied or marked resolved.Suggestions cannot be applied from pending reviews.Suggestions cannot be applied on multi-line comments.Suggestions cannot be applied while the pull request is queued to merge.Suggestion cannot be applied right now. Please check back later.
Replaces #16 and #17 with the versions under upstream review, apache/datafusion-sqlparser-rs#2614 and #2616, which went further after review:
INTERVALis a column name when no literal follows (SELECT interval FROM t,interval - 1), a double-quoted string is an interval value, a bare number still requires a unit, and a string window frame bound stays a string literal.Reverts #17 and #16 on top of #18, then cherry-picks the six upstream commits unchanged (four from apache#2614, two from apache#2616). The
src/and test changes are identical, line for line, to the upstream PRs; only conflict context in imports and test-file tails differs. The fork's full suite (--all-features), fmt and clippy are clean.🤖 Generated with Claude Code
Note
Medium Risk
Changes expression parsing and AST shape for Spark/Databricks intervals and reserved-word fallback; other dialects are mostly unaffected but shared parser paths were touched.
Overview
Aligns Spark and Databricks interval parsing with upstream sqlparser behavior via a new dialect flag,
supports_interval_string_without_qualifier.For those dialects,
INTERVAL '2 months'(and similar string literals, including double-quoted values) can omit an explicit unit; bareINTERVAL 1andINTERVAL 1 + 1 DAYare rejected, while numeric intervals with an explicit unit still parse.INTERVALis treated as a column name when it is not immediately followed by a number or string literal (e.g.SELECT interval FROM t,interval - 1).Core parser updates route interval values through
parse_interval_literal_value, tighten when a unit qualifier is required, and stop treating quoted window frame bounds like'1 day'as interval literals on these dialects. Shared tests cover the new rules; Spark and Databricks suites add targeted interval cases.Reviewed by Cursor Bugbot for commit 485484d. Bugbot is set up for automated code reviews on this repo. Configure here.