Yang Jie created SPARK-58027:
--------------------------------
Summary: Avoid unnecessary Matcher allocation in
SparkDateTimeUtils.getZoneId for region-based zone IDs
Key: SPARK-58027
URL: https://issues.apache.org/jira/browse/SPARK-58027
Project: Spark
Issue Type: Improvement
Components: SQL
Affects Versions: 5.0.0
Reporter: Yang Jie
`SparkDateTimeUtils.getZoneId` runs two regex `replaceFirst` calls on every
invocation to normalize the legacy `(+|-)h:mm` and `(+|-)hh:m` zone-offset
formats (supported before Spark 3.0):
{code:scala}
var formattedZoneId = singleHourTz.matcher(timeZoneId).replaceFirst("$10$2:")
formattedZoneId =
singleMinuteTz.matcher(formattedZoneId).replaceFirst("$1$2:0$3")
ZoneId.of(formattedZoneId, ZoneId.SHORT_IDS)
{code}
Both patterns are anchored on a leading sign character `(+|-)`. Region-based
IDs such as `UTC`, `America/New_York`, or `GMT+8` -- the overwhelming majority
of inputs -- can never match, yet each call still allocates two `Matcher`
instances and scans the string twice before `ZoneId.of` does the real work.
This guards the normalization so the two regexes only run when the input starts
with a sign character, skipping the `Matcher` allocations for the common
region-id case. The change is behavior-preserving: since both patterns are
anchored on the leading sign, an input that does not start with `+`/`-` is
returned unchanged by `replaceFirst` today, which is exactly what the guarded
`else` branch returns.
This is a small clarity and allocation cleanup on the shared `getZoneId`
helper, not a headline performance change.
--
This message was sent by Atlassian Jira
(v8.20.10#820010)
---------------------------------------------------------------------
To unsubscribe, e-mail: [email protected]
For additional commands, e-mail: [email protected]
