ianmcook commented on code in PR #50146:
URL: https://github.com/apache/arrow/pull/50146#discussion_r4201271196
##########
cpp/src/arrow/csv/converter.cc:
##########
@@ -464,18 +472,93 @@ struct MultipleParsersTimestampValueDecoder : public
ValueDecoder {
}
protected:
- using ParserVector = std::vector<const TimestampParser*>;
+ TimeUnit::type unit_;
+ bool expect_timezone_;
+ std::vector<const TimestampParser*> parsers_;
+};
- static ParserVector GetParsers(const ConvertOptions& options) {
- ParserVector parsers(options.timestamp_parsers.size());
- for (size_t i = 0; i < options.timestamp_parsers.size(); ++i) {
- parsers[i] = options.timestamp_parsers[i].get();
+//
+// Value decoder for dates and times, with fallback to user-defined
+// timestamp parsers
+//
+
+// Tries the ISO-8601 format first, then the user-defined timestamp parsers.
+// A timestamp produced by a user-defined parser is floored to the day
+// boundary for dates, and reduced to the time of day for times (consistent
+// with casting a timestamp to date32/date64/time32/time64).
+template <typename T>
+struct DateTimeWithParsersValueDecoder : public ValueDecoder {
+ using value_type = typename T::c_type;
+
+ DateTimeWithParsersValueDecoder(const std::shared_ptr<DataType>& type,
+ const ConvertOptions& options,
+ const TrieCache* trie_cache)
+ : ValueDecoder(type, options, trie_cache),
+ concrete_type_(checked_cast<const T&>(*type)),
+ parse_unit_(GetParseUnit(concrete_type_)),
+ ticks_per_day_(TicksPerDay(parse_unit_)),
+ parsers_(GetTimestampParsers(options_)) {}
+
+ Status Decode(const uint8_t* data, uint32_t size, bool quoted, value_type*
out) {
+ TrimWhiteSpace(&data, &size);
+ if (ARROW_PREDICT_TRUE(string_converter_.Convert(
+ concrete_type_, reinterpret_cast<const char*>(data), size, out))) {
+ return Status::OK();
+ }
+ for (const auto& parser : parsers_) {
+ int64_t timestamp = 0;
+ bool zone_offset_present = false;
+ if (parser->operator()(reinterpret_cast<const char*>(data), size,
parse_unit_,
+ ×tamp, &zone_offset_present) &&
+ !zone_offset_present) {
+ // Floor division, to handle values before the epoch
+ int64_t days = timestamp / ticks_per_day_;
+ days -= (timestamp % ticks_per_day_) < 0;
+ if constexpr (std::is_same_v<T, Date32Type>) {
+ *out = static_cast<value_type>(days);
+ } else if constexpr (std::is_same_v<T, Date64Type>) {
+ *out = days * kMillisPerDay;
+ } else {
+ static_assert(is_time_type<T>::value);
+ *out = static_cast<value_type>(timestamp - days * ticks_per_day_);
Review Comment:
The multiplication `days * ticks_per_day_` can overflow even when the parsed
timestamp and resulting time of day are both representable.
For example, parsing `1677-09-21 00:12:44` into `time64[ns]` with
`timestamp_parsers=[ISO8601]` succeeds at the timestamp-parsing step, but UBSan
then reports:
```text
signed integer overflow: -106752 * 86400000000000 cannot be represented in
type 'int64_t'
```
Could we compute the time of day using a normalized remainder instead?
```cpp
int64_t time_of_day = timestamp % ticks_per_day_;
if (time_of_day < 0) {
time_of_day += ticks_per_day_;
}
*out = static_cast<value_type>(time_of_day);
```
This preserves the intended behavior for pre-epoch timestamps without the
overflowing intermediate multiplication. A regression test using the input
above would cover this boundary.
--
This is an automated message from the Apache Git Service.
To respond to the message, please log on to GitHub and use the
URL above to go to the specific comment.
To unsubscribe, e-mail: [email protected]
For queries about this service, please contact Infrastructure at:
[email protected]
