[lqiu96]

Summary & Bottleneck

BigQuery's REST API encodes INTERVAL values in canonical "[sign]Y-M [sign]D [sign]H:M:S[.F]" format (for example, "123-7 -19 0:24:12.000006"). Previously, FieldValue.getPeriodDuration() and FieldValue.parseCanonicalInterval(String) suffered from three per-cell bottlenecks ($O(\text{rows} \times \text{interval_cols})$):

• getPeriodDuration() unconditionally called PeriodDuration.parse(getStringValue()) inside a try block first. Because ISO-8601 durations always start with 'P' (optionally preceded by '+' or '-'), 100% of canonical BigQuery INTERVAL cells threw and populated a DateTimeParseException stack trace on the normal execution path before falling back to parseCanonicalInterval.
• parseCanonicalInterval(String) compiled a 10-group named-capture regular expression (Pattern.compile(...)) from scratch as a local variable on every cell invocation.
• After extracting hours, minutes, seconds, and fraction from the regex matcher, parseCanonicalInterval concatenated a synthetic "PT...H...M...S" string and parsed the numbers a second time via Duration.parse(durationString).

Changes

• In java-bigquery/google-cloud-bigquery/src/main/java/com/google/cloud/bigquery/FieldValue.java:
  • Hoist the canonical interval regular expression to a private static final Pattern CANONICAL_INTERVAL_PATTERN constant.
  • Add isIso8601PeriodDuration(String) in getPeriodDuration() so PeriodDuration.parse is only attempted when the string begins with 'P' or 'p' (optionally preceded by '+' or '-'), dispatching directly to parseCanonicalInterval for canonical interval strings without throwing DateTimeParseException.
  • Compute Duration.ofSeconds(totalSeconds, nanos) directly from the parsed hours, minutes, seconds, and fraction groups instead of concatenating a "PT..." string and calling Duration.parse.
• In java-bigquery/google-cloud-bigquery/src/test/java/com/google/cloud/bigquery/FieldValueTest.java:
  • Expand testParseCanonicalInterval to test getPeriodDuration() across canonical intervals, signed ISO-8601 intervals, and invalid strings.

Expected Impact

• Eliminates 1 DateTimeParseException stack-trace capture per canonical INTERVAL cell.
• Eliminates 1 full 10-group regex Pattern.compile per INTERVAL cell.
• Eliminates 1 "PT..." string concatenation and Duration.parse call per INTERVAL cell.

Verification

• Ran unit tests in FieldValueTest and FieldValueListTest:
  mvn test -pl java-bigquery/google-cloud-bigquery -Dtest=FieldValueTest,FieldValueListTest