Unexpected results near DST 2023 boundaries (Europe/Helsinki)

[pulakkam]

Hello,

I checked if there are any problems related to DST 2023 in Europe/Helsinki and there seem to be some.

My test code:

import spacetime from 'spacetime'

const isotimes = [
  '2023-03-26T00:30:00.000Z', // March 2023
  '2023-03-26T01:30:00.000Z',
  '2023-03-26T02:30:00.000Z',
  '2023-03-26T03:30:00.000Z',
  '2023-03-26T04:30:00.000Z',
  '2023-03-26T05:30:00.000Z',
  '2023-10-29T00:30:00.000Z', // October 2023
  '2023-10-29T01:30:00.000Z',
  '2023-10-29T02:30:00.000Z',
  '2023-10-29T03:30:00.000Z',
  '2023-10-29T04:30:00.000Z',
]

isotimes.forEach((isotime) => {
  const time = spacetime(isotime)
  console.log(isotime)
  console.log(time.format('iso'))
  console.log(time.goto('Europe/Helsinki').format('iso'))
  console.log()
})

Results:

2023-03-26T00:30:00.000Z
2023-03-26T00:30:00.000Z
2023-03-26T02:30:00.000+02:00

2023-03-26T01:30:00.000Z
2023-03-26T01:30:00.000Z
2023-03-26T05:30:00.000+03:00 // Expected: 2023-03-26T04:30:00.000+03:00

2023-03-26T02:30:00.000Z
2023-03-26T02:30:00.000Z
2023-03-26T06:30:00.000+03:00 // Expected: 2023-03-26T05:30:00.000+03:00

2023-03-26T03:30:00.000Z
2023-03-26T02:30:00.000Z      // Expected: 2023-03-26T03:30:00.000Z
2023-03-26T06:30:00.000+03:00

2023-03-26T04:30:00.000Z
2023-03-26T04:30:00.000Z
2023-03-26T07:30:00.000+03:00

2023-03-26T05:30:00.000Z
2023-03-26T05:30:00.000Z
2023-03-26T08:30:00.000+03:00

2023-10-29T00:30:00.000Z
2023-10-29T00:30:00.000Z
2023-10-29T03:30:00.000+03:00

2023-10-29T01:30:00.000Z
2023-10-29T00:30:00.000Z      // Expected: 2023-10-29T01:30:00.000Z
2023-10-29T03:30:00.000+03:00

2023-10-29T02:30:00.000Z
2023-10-29T02:30:00.000Z
2023-10-29T03:30:00.000+02:00 // Expected: 2023-10-29T04:30:00.000+02:00

2023-10-29T03:30:00.000Z
2023-10-29T03:30:00.000Z
2023-10-29T05:30:00.000+02:00

2023-10-29T04:30:00.000Z
2023-10-29T04:30:00.000Z
2023-10-29T06:30:00.000+02:00

Is there something I could do to get expected results?

Thank you.

[spencermountain]

hey Matti - thank you for the good bug. I think i know what is happening, and it is crazy.

There's some ambiguity when parsing a iso timezone. We get '+05:30' and we try to match it to the most appropriate iana code. We let users clear this up with spacetime(iso, tz).
There are cases when those two timezones don't line up - and we flail around trying to guess timezones.

I think that's what's happening here.

can you confirm that this produces the correct results?

isotimes.forEach((isotime) => {
  const time = spacetime(isotime, 'ETC/UTC')
  console.log(isotime)
  console.log(time.format('iso'))
  console.log(time.goto('Europe/Helsinki').format('iso'))
  console.log()
})

This is pretty dirty. I'll need to think about what the best way to address this is.
let me know if that clears it up
cheers

[pulakkam]

Actually the results are identical, no difference between my original code and your version.

[spencermountain]

oof, sorry! lemme check it out after a good sleep.

[spencermountain]

oh - yup, this is a dupe of #235
sorry for my code-blindness yeterday
i'm working on a fix. it requires a re-write. Hope to have it ready in a couple weeks, atleast as a beta
thanks for your patience. the offsets will only be off by an hour, within an hour of a dst change
cheers