Unexpected behavior of "by=group" in combination with chron()

Start by running the reproducible data.table and chron example from the issue and compare the grouped chron results with paste(). Trace the grouped evaluation path implicated by the verbose output. Done means each yyyy group produces chron timestamps using its own year, with a regression test covering the reported case.

First off, thank you so very much for the amazing data.table package, which I have been using almost daily for many years.

I searched data.table NEWS and stack overflow [data.table] for previous mentions of this issue and did not find any.

There appears to be a data.table bug related to the use of "by=group" in combination with chron() from the chron package. In the following reproducible example, data.table with "by=group" seems to call chron() in such a way that information from the first group is leaked to and used by subsequent groups, unlike the way it works for other functions (I used paste() in the example). I do not think this is a chron bug, because data from the first "by=group" should not be available to subsequent groups, correct?

Notes: The data.table foo contains time/date information encoded as seconds since the beginning of the year (time.s) and year (yyyy). Chron can assemble this information into a timestamp via chron(time.s/60/60/24,origin=c(1,1,yyyy[1])). With "group=yyyy", the chron origin and resulting timestamp should vary by yyyy group, but they do not. The desired behavior is for each timestamp to use the correct year, e.g.: (01/01/19 00:01:01) instead of (01/01/17 00:01:01).

I have tried many variants of this code (e.g., re-ordering the groups, grouping by a third variable, or omitting a subset of lines). In all cases, all chron timestamps wound up using the year from the first group, as if "by=yyyy" had been omitted, whereas paste(yyyy[1],time.s) works as desired and expected.

I got the same behavior on three different machines and R versions as far back as 3.6.1. The example session info is from an instance of RStudio Cloud.

Thanks for your help,

-Klaus Huebert

library(data.table)
library(chron)
options(datatable.verbose=TRUE)

foo<-data.table(time.s=59:62, yyyy=c(2017,2019,2018,2017))

foo
#    time.s yyyy
# 1:     59 2017
# 2:     60 2019
# 3:     61 2018
# 4:     62 2017

foo[,.(paste=paste(yyyy[1], time.s), chron=chron(time.s/60/60/24, origin=c(1, 1, yyyy[1]))), by=yyyy]
# Detected that j uses these columns: time.s 
# Finding groups using forderv ... forder.c received 4 rows and 1 columns
# 0.000s elapsed (0.000s cpu) 
# Finding group sizes from the positions (can be avoided to save RAM) ... 0.000s elapsed (0.000s cpu) 
# Getting back original order ... forder.c received a vector type 'integer' length 3
# 0.000s elapsed (0.000s cpu) 
# lapply optimization is on, j unchanged as 'list(paste(yyyy[1], time.s), chron(time.s/60/60/24, origin = c(1, 1, yyyy[1])))'
# GForce is on, left j unchanged
# Old mean optimization is on, left j unchanged.
# Making each group and running j (GForce FALSE) ... 
#   collecting discontiguous groups took 0.000s for 3 groups
#   eval(j) took 0.000s for 3 calls
# 0.000s elapsed (0.000s cpu) 
#    yyyy   paste               chron
# 1: 2017 2017 59 (01/01/17 00:00:59)
# 2: 2017 2017 62 (01/01/17 00:01:02)
# 3: 2019 2019 60 (01/01/17 00:01:00)
# 4: 2018 2018 61 (01/01/17 00:01:01)

sessionInfo()
# R version 4.2.1 (2022-06-23)
# Platform: x86_64-pc-linux-gnu (64-bit)
# Running under: Ubuntu 20.04.5 LTS
# 
# Matrix products: default
# BLAS:   /usr/lib/x86_64-linux-gnu/atlas/libblas.so.3.10.3
# LAPACK: /usr/lib/x86_64-linux-gnu/atlas/liblapack.so.3.10.3
# 
# locale:
#  [1] LC_CTYPE=C.UTF-8       LC_NUMERIC=C           LC_TIME=C.UTF-8        LC_COLLATE=C.UTF-8    
#  [5] LC_MONETARY=C.UTF-8    LC_MESSAGES=C.UTF-8    LC_PAPER=C.UTF-8       LC_NAME=C             
#  [9] LC_ADDRESS=C           LC_TELEPHONE=C         LC_MEASUREMENT=C.UTF-8 LC_IDENTIFICATION=C   
# 
# attached base packages:
# [1] stats     graphics  grDevices utils     datasets  methods   base     
# 
# other attached packages:
# [1] chron_2.3-58      data.table_1.14.2
# 
# loaded via a namespace (and not attached):
# [1] compiler_4.2.1 tools_4.2.1