calculate_player_stats_def() no longer errors with small subsets of pbp

[mrcaseb]

if the pbp subset is small, there could be several columns missing which results in dplyr errors.

This PR fixes that behavior by adding the relevant columns if they are missing

closes #410

[mrcaseb]

FYI: I checked the function with various random samples of 2021 pbp. I think it's robust now

pbp <- nflfastR::load_pbp(2021)

pbp %>% 
  dplyr::slice_sample(n = 1) %>% 
  calculate_player_stats_def()

[tanho63]

I solved the problem a bit differently here, what approach do we think is better? https://github.com/nflverse/ngsscrapR/blob/master/R/roster.R#L176

[mrcaseb]

I like both! In terms of the playstats function I prefer having exact control over what I add and avoid additional rows or new NAs at the same time.

[tanho63]

I like the other approach for having better control over vector types (and they’re all length zero so no new rows).

maybe if this accepted named args of empty vectors eg

add_column_if_missing(df, forced_fumble_player = character(), n_fumble_forced = numeric()) ?

[mrcaseb]

Yeah that's easy to implement

[mrcaseb]

I wonder if it makes sense in this case tho. We would exclusively do integer as all of the relevant columns are only counts

[tanho63]

player is an ID that should be character?

[mrcaseb]

I may have missed something. Gotta check later

[mrcaseb]

Double checked and all relevant variables are actually integers

"solo_tackle", 
"tackle_with_assist", 
"tackle_for_loss", 
"assist_tackle",
"forced_fumble_player" # sounds like player ID but in fact is a counter,
"n_sack", 
"n_qb_hit", 
"sack_yards_sack",
"n_interception", 
"n_pass_defense", 
"return_yards_interception",
"fumble_recovery",
"n_penalty", 
"yards_penalty"

[github-actions]

🚀 Deployed on https://64b8eeb762e102576d715278--thirsty-brown-d38ddf.netlify.app