Tim Armstrong has posted comments on this change. Change subject: IMPALA-3894: Changed the behavior parsing 2-digit year values ......................................................................
Patch Set 4: (8 comments) Looks pretty good. No major concerns, we just need some tests and a bit of cleanup to make it easier to follow. http://gerrit.cloudera.org:8080/#/c/7530/4//COMMIT_MSG Commit Message: Line 16: > select from_unixtime(unix_timestamp('31-AUG-94', 'dd-MMM-yy'),'yyyyMMdd'); Can you add some tests in expr-test for some different cases to cover all of the code paths? I think you can make the value of "now()" deterministic if you set up a RuntimeState(), and pass in a TQueryCtx with now_string set. http://gerrit.cloudera.org:8080/#/c/7530/4/be/src/exprs/timestamp-functions-ir.cc File be/src/exprs/timestamp-functions-ir.cc: Line 159: dt_ctx->now = context->impl()->state()->now(); We should be able to do this during the Prepare() phase of this function. It looks like that's implemented in UnixAndFromUnixPrepare() according to the function registry: https://github.com/apache/incubator-impala/blob/master/common/function-registry/impala_functions.py#L246 http://gerrit.cloudera.org:8080/#/c/7530/4/be/src/runtime/timestamp-parse-util.cc File be/src/runtime/timestamp-parse-util.cc: Line 449: if (tok_len <= 2) { You can do this on one line (see https://google.github.io/styleguide/cppguide.html#Conditionals) if (tok_len <= 2) realign_year = true; PS4, Line 549: dt_ctx.now->date() Let's factor this common subexpression out into a separate variable for readability. Line 551: dt_result->year += (century_start_year / 100) * 100; Could you add a comment giving an example of what the value might be at this point. It's subtle.. Line 555: if (TimestampValue(parsed_date, parsed_time) < Could you add a comment giving an example of when we need to increment by 100? http://gerrit.cloudera.org:8080/#/c/7530/4/be/src/runtime/timestamp-parse-util.h File be/src/runtime/timestamp-parse-util.h: Line 139: /// Current time to determine the actual year when parsing 2-digit year token nit: add a "." for consistency with other comments. Line 140: const TimestampValue* now; Let's just store a copy of the TimestampValue inline instead of the pointer. Makes it easier to understand who owns the memory. -- To view, visit http://gerrit.cloudera.org:8080/7530 To unsubscribe, visit http://gerrit.cloudera.org:8080/settings Gerrit-MessageType: comment Gerrit-Change-Id: I5da761255915dc741f1dcc488fd4ef6ecc385896 Gerrit-PatchSet: 4 Gerrit-Project: Impala-ASF Gerrit-Branch: master Gerrit-Owner: Tianyi Wang <[email protected]> Gerrit-Reviewer: Greg Rahn <[email protected]> Gerrit-Reviewer: Tim Armstrong <[email protected]> Gerrit-HasComments: Yes
