[ARROW-10343] [C++] Unable to parse strings into timestamps

• Type: Bug
• Status: Closed
• Priority: Minor
• Resolution: Duplicate
• Affects Version/s: 1.0.1
• Component/s: C++, Python
• Labels: timestamp timezone
• Environment: macOS 10.15.7, Python 3.8.2
• External issue URL: https://github.com/apache/arrow/issues/26330
• Assignee: Unassigned
• Reporter: Niclas Roos
• Created: 19/Oct/20 07:43
• Updated: 11/Jan/23 08:12
• Resolved: 23/Nov/21 19:20

Description

Hi,

I'm working with parquet files generated by a AWS RDS Postgres snapshot export. 

I'm trying to parse a date column stored as a string into a timestamp, but it fails.

I've managed to parse the same date format (as in the first example below) when reading from a csv, so I tried to investigate it as far as I could on my own, and here's my results:

import pyarrow as pa
import pytz

#################################################################################
## the format I get from the database
us_tz_arr = pa.array([
  "2014-12-07 07:48:59.285332+00",
  "2014-12-07 08:01:49.758975+00",
  "2014-12-07 10:11:35.884304+00"])

us_tz_arr.cast(pa.timestamp('us', tz=pytz.UTC))
-> ArrowInvalid: Failed to parse string: 2014-12-07 10:11:35.884304+00

#################################################################################
## tried removing the timezone
us_arr = pa.array([
  "2014-12-07 07:48:59.285332",
  "2014-12-07 08:01:49.758975",
  "2014-12-07 10:11:35.884304"])

us_arr.cast(pa.timestamp('us'))
-> ArrowInvalid: Failed to parse string: 2014-12-07 10:11:35.884304

#################################################################################
## tried removing the microseconds but keeping the timezone
second_tz_arr = pa.array([
  "2014-12-07 07:48:59+00",
  "2014-12-07 08:01:49+00",
  "2014-12-07 10:11:35+00"])

second_tz_arr.cast(pa.timestamp('s', tz=pytz.UTC))
-> ArrowInvalid: Failed to parse string: 2014-12-07 10:11:35+00

#################################################################################
## removing microseconds and timezone, makes it work!
s_arr = pa.array([
  "2014-12-07 07:48:59",
  "2014-12-07 08:01:49",
  "2014-12-07 10:11:35"])

s_arr.cast(pa.timestamp('s'))
-> <pyarrow.lib.TimestampArray object at 0x7fbdf81ae460>
[
  2014-12-07 07:48:59,
  2014-12-07 08:01:49,
  2014-12-07 10:11:35
]

 PS. This is my first bug report, so apologies if important things are missing.

Issue Links

• is fixed by: ARROW-12820 [C++] Strptime ignores timezone information
• is related to: ARROW-13625 [C++][CSV] Timestamp parsing should accept any valid ISO 8601 without requiring custom parse strings