Github user majetideepak commented on a diff in the pull request:
https://github.com/apache/orc/pull/110#discussion_r114200187
--- Diff: c++/src/Statistics.cc ---
@@ -279,20 +277,55 @@ namespace orc {
}
TimestampColumnStatisticsImpl::TimestampColumnStatisticsImpl
- (const proto::ColumnStatistics& pb, bool correctStats) {
+ (const proto::ColumnStatistics& pb, const StatContext& statContext) {
valueCount = pb.numberofvalues();
- if (!pb.has_timestampstatistics() || !correctStats) {
+ if (!pb.has_timestampstatistics() || !statContext.correctStats) {
_hasMinimum = false;
_hasMaximum = false;
+ _hasLowerBound = false;
+ _hasUpperBound = false;
minimum = 0;
maximum = 0;
+ lowerBound = 0;
+ upperBound = 0;
}else{
const proto::TimestampStatistics& stats = pb.timestampstatistics();
- _hasMinimum = stats.has_minimum();
- _hasMaximum = stats.has_maximum();
-
- minimum = stats.minimum();
- maximum = stats.maximum();
+ _hasMinimum = stats.has_minimumutc() || (stats.has_minimum() &&
(statContext.writerTimezone != NULL));
+ _hasMaximum = stats.has_maximumutc() || (stats.has_maximum() &&
(statContext.writerTimezone != NULL));
+ _hasLowerBound = stats.has_minimumutc() || stats.has_minimum();
+ _hasUpperBound = stats.has_maximumutc() || stats.has_maximum();
+
+ // Timestamp stats are stored in milliseconds
+ if (stats.has_minimumutc()) {
+ minimum = stats.minimumutc();
+ lowerBound = minimum;
+ } else if (statContext.writerTimezone) {
+ int64_t writerTimeSec = stats.minimum() / 1000;
+ // multiply the offset by 1000 to convert to millisecond
+ minimum = stats.minimum() +
(statContext.writerTimezone->getVariant(writerTimeSec).gmtOffset) * 1000;
+ lowerBound = minimum;
+ } else {
+ minimum = 0;
+ // subtract 1 day to handle unknown TZ
+ lowerBound = stats.minimum() - (24 * 60 * 60 * 1000);
+ }
+
+ // Timestamp stats are stored in milliseconds
+ if (stats.has_maximumutc()) {
+ maximum = stats.maximumutc();
+ upperBound = maximum;
+ } else if (statContext.writerTimezone) {
+ int64_t writerTimeSec = stats.maximum() / 1000;
+ // multiply the offset by 1000 to convert to millisecond
+ maximum = stats.maximum() +
(statContext.writerTimezone->getVariant(writerTimeSec).gmtOffset) * 1000;
+ upperBound = maximum;
+ } else {
+ maximum = 0;
+ // add 1 day to handle unknown TZ
+ upperBound = stats.maximum() + (SECONDS_PER_DAY * 1000);
+ }
+ // Add 1 millisecond to account for microsecond precision of values
+ maximum += 1;
--- End diff --
The timestamp values have microsecond precision `eg: 2037-01-01
00:00:00.000999`.
However, I see that the statistics computed have millisecond precision
`2037-01-01 00:00:00.000` and they are not rounded off to the ceil. This is to
account for the ceil. I will update the test.
Thanks! for catching this.
---
If your project is set up for it, you can reply to this email and have your
reply appear on GitHub as well. If your project does not have this feature
enabled and wishes so, or if the feature is enabled but not working, please
contact infrastructure at [email protected] or file a JIRA ticket
with INFRA.
---
