{
  "$type": "site.standard.document",
  "bskyPostRef": {
    "cid": "bafyreigfznymlerzwmr2dse3hhiuuypivkdpt6gbfk3565oqycg5cc4mgu",
    "uri": "at://did:plc:pi6woz4d47bkuws673w2il2r/app.bsky.feed.post/3mhqmorskma52"
  },
  "path": "/t/rfc-sibyl-time-series-analysis-in-haskell/13823#post_20",
  "publishedAt": "2026-03-22T22:52:14.000Z",
  "site": "https://discourse.haskell.org",
  "tags": [
    "@mchav"
  ],
  "textContent": "Thanks for looking at the library - originally I figured we might do something like a “Frequency” type (Monthly, Daily, Hourly, Minutely, Secondly, etc) but decided against it because I felt it was too rigid. I’m a student who works with time series in the very pedantic sense, and so insight into common use cases for statisticians would be really really useful.\n\nHow would you suggest we represent the type internally? An issue we ran into also was rather to have allow nulls in data, or if we should make the user clean up their data before asking Sibyl for help, which I’m starting to think isn’t the correct approach. It may end up that Sibyl operates on DataFrames themselves, and allows for irregular spaced time data and other unordered data, or intervals of data. Working out how to represent that in Haskell has become a challenge for me during planning… also @mchav what do you think?",
  "title": "[RFC] Sibyl: Time Series Analysis in Haskell"
}