.mom files, <mom> tibbles

Source: vignettes/ii_mom_mom.Rmd
Work in progress

> a common file format is essential for you and morphometrics
> .mom files and <mom> tibbles are the same object, presented differently

Rationale for a common morphometric format

Since I started morphometrics on a daily basis, I have always been surprised that a common file format for morphometrics, to the best of my knowledge, does not exist.

Perhaps to my bigger surprise, it even does not seem to be really missing.

A common file format have direct benefits from the single user to the community scale: it is easy to share, archive, reuse, recycle, publish, etc.

In my humble opinion, morphometrics as a whole, and the questions it feeds, can hardly scale up without a common file format.

.mom : yaml file and markup language

An ideal file format for morphometrics would be:

  • easy to read for humans
  • easy to parse for machines
  • easy to write and develop on

Good news, this exists : yaml (todo) (and more generally…). yaml and structure language wikipedia’s pages are nice ones.

For general i/o operations with yaml, the eponym package by xx is just perfect.

Relations between <mom> and .mom

A .mom file is nothing but a yaml file with the .mom extension. Actually, .mom files are textual and strutured representation of a <mom> tibble, no matter how small or complex it is.

Chivas looks like this in R, eg as a <mom> tibble:

library(Momit)
# shorten it for the sake of readability
chivas <- chivas %>% Momocs2::coo_sample(4)

And like this in yaml, eg in a .mom:

chivas_yaml <- chivas %>% yaml::as.yaml()
# cat help visualize what would be written in a a .mom file
chivas_yaml %>% cat()
#> coo:
#>   chivas:
#>     x:
#>     - 33.0
#>     - 141.0
#>     - 329.0
#>     - 238.0
#>     'y':
#>     - 437.0
#>     - 44.0
#>     - 201.0
#>     - 668.0
#> type: beer
#> fake: c
#> size: -123.45
#> missing: .na

Now, let’s chivas_yaml is read back into R:

chivas2 <- chivas_yaml %>% yaml::yaml.load() %>% tibble::as_tibble()
chivas2
#> # A tibble: 1 x 5
#>   coo              type  fake   size missing
#>   <named list>     <chr> <chr> <dbl> <lgl>  
#> 1 <named list [2]> beer  c     -123. NA

And, finally, let’s compare with the original tibble. We unsophisticate the mom for a tibble:

chivas %>% tibble::as_tibble()
#> # A tibble: 1 x 5
#>   coo                    type  fake   size missing
#>   <list<coo_single[,2]>> <fct> <fct> <dbl> <lgl>  
#> 1 <tibble [4 × 2]>       beer  c     -123. NA

Not bad isn’t? If you dissecate the two prints, you will notice that three things were dropped:

  1. the mom_df class (but we asked for it)
  2. the coo_list nature of the coo column. Now it’s just a list
  3. the names within this column