readr::read_csv excludes milliseconds from timestamps

[bastianilso]

(see original issue at tidyverse/purrr#764)

reading the following csv file with readr::read_csv removes milliseconds in timestamps (formatted HH:mm:ss.ffff). Reading with base read.csv works fine.

with the following file
sampleLog_2020-05-13 10-30-10.0167.zip

try:

  df2<-readr::read_csv("sampleLog_2020-05-13 10-30-10.0167.csv")

compare to

  df2<-read.csv("sampleLog_2020-05-13 10-30-10.0167.csv")

[jimhester]

This is a duplicate of #883

readr is actually reading the seconds, but R does not display fractional seconds by default.

library(readr)
data <- read_csv("~/p/readr/sampleLog_2020-05-13 10-30-10.0167.csv")
#> Parsed with column specification:
#> cols(
#>   .default = col_double(),
#>   TimeStamp = col_datetime(format = ""),
#>   Time = col_time(format = ""),
#>   Date = col_date(format = ""),
#>   Event = col_character(),
#>   PupilTime = col_character(),
#>   UnityToPupilTimeOffset = col_character(),
#>   GazeConfidence = col_character(),
#>   EyeCenter0X = col_character(),
#>   EyeCenter0Y = col_character(),
#>   EyeCenter0Z = col_character(),
#>   EyeCenter1X = col_character(),
#>   EyeCenter1Y = col_character(),
#>   EyeCenter1Z = col_character(),
#>   GazeNormal0X = col_character(),
#>   GazeNormal0Y = col_character(),
#>   GazeNormal0Z = col_character(),
#>   GazeNormal1X = col_character(),
#>   GazeNormal1Y = col_character(),
#>   GazeNormal1Z = col_character(),
#>   LocalGazeDirectionX = col_character()
#>   # ... with 17 more columns
#> )
#> See spec(...) for full column specifications.

as.POSIXlt(data$TimeStamp[[1]])$sec
#> [1] 10.0177

^{Created on 2020-06-05 by the reprex package (v0.3.0)}

[bastianilso]

@jimhester I tried your command below and it works for the column Timestamp. But when I try to do it on the column Time, I again only get the result in seconds.

[kalenkovich]

The issue is still present 2.1.4.

@jimhester, I don't think this is the display issue: OP's problem (and mine) is with the Time column, not TimeStamp. readr correctly guesses that the Time column contains time values and parses it as "hms" but the milliseconds are silently dropped.

library(readr)

filename <- "sampleLog_2020-05-13 10-30-10.0167.csv"
test <- read_csv(filename)
print(as.double(test$Time[[1]]), digits = 20)
# 37810

test2 <- read_csv(filename, col_types = cols(Time = col_character))
> print(as.double(as_hms(test2$Time[[1]])), digits = 20)
[1] 37810.017699956893921

Maybe there is a format specification for col_time that would work here? But in any case, it would be useful if readr at least produced a warning about the loss of precision.