[magnusp]

Fixes #201

Extracts storage-class to date/time conversion from LibsqlResultSet into package-private SqliteDateTimeValues (from(Value, String targetType) plus epochMillisToLocalDateTime, julianDayToEpochMillis, parseSqliteDateTime, parseSqliteTime). A thin dateTimeValue(int, String) stays in the cursor layer: it still calls getCurrentValue(columnIndex) (which sets lastColumnIndex, so wasNull keeps working) and then delegates. getLocalDate, getLocalTime, and getLocalDateTime are unchanged.

Moved verbatim (normalized diff)

Compared the parent block (LibsqlResultSet.java lines 795-895 at 373b875: the switch body plus the four helpers) with the new class (from body plus helpers), normalizing only by dropping private from the four helper signatures and trimming trailing whitespace:

$ sed -E 's/\bprivate //g; s/[[:space:]]+$//' old_moved.txt > old_norm.txt
$ sed -E 's/\bprivate //g; s/[[:space:]]+$//' new_moved.txt > new_norm.txt
$ diff -u old_norm.txt new_norm.txt && echo "NORMALIZED DIFF: EMPTY"
NORMALIZED DIFF: EMPTY
$ sha256sum old_norm.txt new_norm.txt
53cfd1c36f62b74b9140c05100f017a4828570327ab79c25e315d911c49f4f50  old_norm.txt
53cfd1c36f62b74b9140c05100f017a4828570327ab79c25e315d911c49f4f50  new_norm.txt

No parse rule, default, message, or zone changed. BLOB still throws "Bad value for type " + targetType; blank text still returns null; bare time still defaults to 2000-01-01; offsets are honored only when the sign index exceeds nine; Z/z is UTC; conversion uses ZoneId.systemDefault(); exception wrapping keeps the original text; INTEGER epoch millis and REAL Julian day are untouched. No dependency, pom.xml, test, numeric-text, or LOB change.

Probe

Scratch edit changed the bare-time default in the moved code from LocalDate.of(2000, 1, 1) to LocalDate.of(1970, 1, 1):

LibsqlResultSetTest.bareTimeTextGetsTheYear2000DefaultDate -- FAILURE!
expected: 2000-01-01T10:30:45 (java.time.LocalDateTime)
 but was: 1970-01-01T10:30:45 (java.time.LocalDateTime)
	at ...LibsqlResultSetTest.bareTimeTextGetsTheYear2000DefaultDate(LibsqlResultSetTest.java:352)

Reverted. The new file's sha256 returned to the pre-probe value 80f18b5c…, git status --porcelain lists only the two intended paths, and git diff -- libsql-jdbc/src/test is empty, so the four wave-0 date/time tests stay green unedited.

Line counts

• LibsqlResultSet.java: 912 → 809 (under 850)
• SqliteDateTimeValues.java: 124 (new)
• Diff stat: 2 files changed, 125 insertions(+), 104 deletions(-)

Lanes

• ./mvnw -pl libsql-jdbc -am test → 392 tests, 0 failures, 0 errors
• ./mvnw spotless:apply, then ./mvnw spotless:check → BUILD SUCCESS both
• ./mvnw -pl libsql-conformance -am verify → 272 tests, 0 failures, 7 skipped

[magnusp]

Root verdict: PASS at fd32836.

Verifier confirmed the moved methods are byte-identical, the four wave-0 date and time tests are unedited and green, and the cursor side effect is intact. A live probe on a NULL column returned null through every date getter with wasNull true, which is the getCurrentValue contract. 427 tests across the reactor and 272 conformance tests green. The bare-time probe bit and was reverted. Countersign: root, fd32836.

This completes wave 1 of the stack.