I tend to set up my targets pipeline in such a way that each target is calling functions written and stored in the typical R folder. All constants, parameter values, file paths, and objects of this nature I tend to declare and manage separate in a file stored in a typical location like inst/scripts/objects.R.
For reference, this is how the pipeline is initialized and all files loaded:
# Load packages required to define the pipeline:
library(targets)
# Set target options:
tar_option_set(
packages = c("magrittr"),
format = "rds" # default storage format
# Set other options as needed.
)
# Source other scripts as needed.
source(here::here('inst', 'scripts', 'objects.R'))
# Set target configuration:
tar_config_set(store = store_path)
# tar_make_clustermq() configuration (okay to leave alone):
options(clustermq.scheduler = "multiprocess")
# Run the R scripts in the R/ folder with your custom functions:
tar_source()
Lately, targets has reported on of these objects stored in objects.R as having changed and thus wanting to re-run my entire pipeline (as it's an object early in the flow). However, there is absolutely zero evidence of this value having changed (I can track the file's history through git). For reference, this value was declared very early in my development phase, and it's as simple as this:
> targets::tar_meta(names = 'germany_code')
# A tibble: 1 × 18
name type data command depend seed path time size bytes format repository iteration parent children seconds warnings error
<chr> <chr> <chr> <chr> <chr> <int> <lis> <dttm> <chr> <int> <chr> <chr> <chr> <lgl> <list> <dbl> <lgl> <chr>
1 germany_code object 92df76c… NA NA NA <chr> NA NA NA NA NA NA NA <chr [1]> NA NA NA
That is it, it's a value of 7, assigned to object germany_code which has never been changed since it was first declared. For additional reference, the values at columns path and children are NA.
I have read all documentation I can find, and especially that relating to the tar_cue() function, which I know is a way to control which targets get re-run:
https://docs.ropensci.org/targets/reference/tar_cue.html
Is there a way to see the history of some of the metadata fields being used to track changes by targets? I'm really struggling to determine why there is change reported when nothing apparent has changed. Are there any other recommendations/tricks I may be missing to try and diagnose what's going on here?
Many thanks in advance for any help you can provide.