athena-express icon indicating copy to clipboard operation
athena-express copied to clipboard

Extend data types support for UTC dates, i.e. `Timestamp` Athena type

Open ssedano opened this issue 3 years ago • 7 comments

Hi team,

I'd like to extend data types support for UTC dates, i.e. Timestamp Athena type. My proposal is to create a new option, e.g. useUtcDates, to cast values of columns with type Timestamp to Date Javascript objects. This setting would have a default value of false. The caveat of this setting is that it will use machine-local time, thus this setting is intended for systems configured in UTC. This could be extended to Date Type but it would transform dates into date and times, e.g. '2020-11-18' would become '2020-11-18T00:00:00Z'. I intentionally avoid casting timestamp with time zone Athena type to reduce complexity.

For applications that deal with dates, converting Timestamp columns to Date objects is a common task given that many libraries expect Date objects. Moreover, athena-express hides the type in the result set. This forces developers that wish to operate using dates to access the column by name and explicitly convert values to Date objects.

I'd be willing to submit a patch if this feature request is approved.

Proposed update to documentation:

  • useUtcDates | boolean | (default false) | Cast Timestamp types to Javascript Date objects with times in UTC. Set this option if your code runs in machines configured with UTC as their timezone. Useful when your Timestamp columns are representing UTC date and time and you want to work with Date objects instead of String objects

ssedano avatar Nov 21 '20 13:11 ssedano

I somehow missed responding to this. If you wanna go ahead and add that, I'll be happy to take a look. Just consider any edge cases that might break it.

ghdna avatar Apr 16 '21 00:04 ghdna

Actually the more I think about this, the more it can cause more confusion for users since there are several date formats to convert into: For e.g. Athena's TIMESTAMP datatype will fetch results as 2008-09-15 03:04:05.327 This string can be converted into 5 formats:

  1. new Date() => Mon Sep 15 2008 03:04:05 GMT-0400 (Eastern Daylight Time)
  2. .toUTCString() => "Mon, 15 Sep 2008 07:04:05 GMT"
  3. .toDateString() => "Mon Sep 15 2008"
  4. .toISOString() => "2008-09-15T07:04:05.327Z"
  5. .toTimeString() => "03:04:05 GMT-0400 (Eastern Daylight Time)"

How would AthenaExpress know which format the user cares about? And the bigger question is, should this be something AthenaExpress does or should this functionality be handed over to the user to decide.

ghdna avatar Apr 25 '21 15:04 ghdna

Thank you for your reply.

should this be something AthenaExpress does or should this functionality be handed over to the user to decide.

Is the right question to ask. Unfortunately, query method doesn't include the data type. Hence why I propose to let the user decide at the time of configuring AthenaExpress. This saves the user from tracking column types separately and casting objects, keeping AthenaExpress' simplicity.

ssedano avatar Apr 27 '21 10:04 ssedano

ok i understand. Let me check on the effort required for this

ghdna avatar Apr 29 '21 11:04 ghdna

I would be happy to contribute the patch

Thank you.

ssedano avatar May 03 '21 11:05 ssedano

Sure, please do. I'd be happy to look at it.

ghdna avatar May 05 '21 04:05 ghdna

Apologies for the delay on this. I expect to have a pull request ready in the next two weeks

ssedano avatar Jun 06 '21 09:06 ssedano