Record doc numbers along with PINs when calculating comps

[jeancochrane]

This PR updates the comps calculation step in the interpret stage to save comp document numbers along with PINs. With document numbers, we can update downstream code that consumes these comps to uniquely tie each comp to a specific sale, instead of having to infer the sale based only on the comp's PIN.

I tested this locally on a subset of training/assessment data. The comps aren't great due to the limited size of the data, but it ran reasonably fast and confirmed that this processing code works. I'm happy to share the output of my test, or to kick off a remote comps run if that would make things easier to review. In the meantime, here's a quick peek at the output schema:

> final_comps <- cbind(comps[[1]], comps[[2]])
> final_comps %>% names
 [1] "pin"            "card"           "comp_pin_1"     "comp_pin_2"     "comp_pin_3"     "comp_pin_4"     "comp_pin_5"     "comp_pin_6"    
 [9] "comp_pin_7"     "comp_pin_8"     "comp_pin_9"     "comp_pin_10"    "comp_pin_11"    "comp_pin_12"    "comp_pin_13"    "comp_pin_14"   
[17] "comp_pin_15"    "comp_pin_16"    "comp_pin_17"    "comp_pin_18"    "comp_pin_19"    "comp_pin_20"    "comp_doc_no_1"  "comp_doc_no_2" 
[25] "comp_doc_no_3"  "comp_doc_no_4"  "comp_doc_no_5"  "comp_doc_no_6"  "comp_doc_no_7"  "comp_doc_no_8"  "comp_doc_no_9"  "comp_doc_no_10"
[33] "comp_doc_no_11" "comp_doc_no_12" "comp_doc_no_13" "comp_doc_no_14" "comp_doc_no_15" "comp_doc_no_16" "comp_doc_no_17" "comp_doc_no_18"
[41] "comp_doc_no_19" "comp_doc_no_20" "comp_score_1"   "comp_score_2"   "comp_score_3"   "comp_score_4"   "comp_score_5"   "comp_score_6"  
[49] "comp_score_7"   "comp_score_8"   "comp_score_9"   "comp_score_10"  "comp_score_11"  "comp_score_12"  "comp_score_13"  "comp_score_14" 
[57] "comp_score_15"  "comp_score_16"  "comp_score_17"  "comp_score_18"  "comp_score_19"  "comp_score_20" 

Closes https://github.com/ccao-data/pinval/issues/7.

[jeancochrane]

There's a variety of different ways we refer to this type of data -- alternately, doc_no or instruno or document_num, depending on the context. What's the best choice for this particular case?

[dfsnow]

Let's stick with what's used in the model pipeline - document_num.

[jeancochrane]

Makes sense, done in d510fe7.

[dfsnow]

Nice, this should be super helpful. Before we merge, let's also use this as an opportunity to test the Batch action by doing a re-run of the final res model with the updated comps. This will give us the new output and will ensure we haven't busted anything in the interim since we wrapped modeling.

[dfsnow]

Let's stick with what's used in the model pipeline - document_num.

[jeancochrane]

@dfsnow I finally got a completed run finished with run ID 2024-06-18-calm-nathan. Want to take a look before I merge?

[dfsnow]

@dfsnow I finally got a completed run finished with run ID 2024-06-18-calm-nathan. Want to take a look before I merge?

@jeancochrane I took a look and everything looks good. Let's merge it!