Problem with dates after the last reticulate update about featuretoolsr HOT 6 CLOSED

Good input, however, coercing to POSIXct also unintentionally adds a time zone to the date, which means the date potentially no longer is the same. It also adds details (hour) that isn't originally there. One way is to coerce to UTC, but I'm not sure it's optimal.

Example (europe):

Sys.Date() %>% as.POSIXct()
## Returns: 2019-04-22 02:00:00 CEST

I think a warning could be a better approach?

from featuretoolsr.

if the user table contains a date column with 10000+ values the R session can be treated as frozen. So, the warning doesn't actually help but definitely is much better than a quiet problem

from featuretoolsr.

I've been thinking about this issue along with a few more of the open ones. I think you're bringing up a fair point, but I don't want to forcibly change user values. On the other hand, if it makes the R session unusable we should adress it forcibly..

What do think about adding support for this through a option-toggle (default on package load: on)?

Something along the lines of:

# Loads on package attach
options(
    featuretoolsR.force_posixct = T, #default
    featuretoolsR.posixct_tz = "UTC" #default
)

Runs something like this on entityset creation:

if(getOption("featuretoolsR.force_posixct")) {
  .cols = lapply(df, class)
  for (i in 1:length(.cols)) {
    colname <- names(.cols)[i]
    coltype <- .cols[[colname]]
    if (any(coltype == "Date")) {
      df[,colname] <- as.POSIXct(df[,colname], getOption("featuretoolsR.posixct_tz"))
    }
  }
}

from featuretoolsr.

I added something like this to the current development-branch (990b40e). Feel free to try it out.

from featuretoolsr.

Sorry, I don't feel I can help here

from featuretoolsr.

Resolved in #17

from featuretoolsr.