Architecture

library(tidyverse)
library(Momocs2)

Let’s have a drink1. Do you prefer beer or wine?

mosaic(bot, type) # add id support

Everything is a tibble

You see the visual representation of a “mom” object that looks like this:

bot                  # shipped with Momocs2

Across MomX, the main object is mom, are tibbles, slightly augmented for morphometrics. Tibbles themselves are built on top of goold old data.frame.

Keep this in mind:

A mom is a tibble, a tibble is a data.frame, a data.frame is a list, and a list can contain anything, including other lists.

Typically, at some point of the MomX pipeline, one or more columns in these tibbles will store coordinates (<coo_list>), coefficients (<coe_list>), etc.

coordinates classes

The coo column above is a <coo_list> a list of single shapes. Let’s have a look:

bot$coo %>% head(2)

Each single shapes are stored in tibbles with exactly 2 columns, named x and y. Such a thing is a <coo_single>.

In morphometrics, shapes rarely come single but as collection of shapes, hence the <coo_list>. Also, they usually have covariates attached that should be firmly tied to it, hence the <mom_tbl>.

coefficients classes

In morphometrics, the destiny of coordinates is to be turned into coefficients. Coefficients are obtained using morphometric methods like elliptical Fourier transforms, or Procrustes alignment.

efourier(bot$coo[[1]])

Here an elliptical Fourier tranform turned a <coo_single> into a <coe_single>. They are tibbles with exactly 1 row but with possiby many columns.

Morphometrics is pivotationnal, or transpositionnal, in nature: it turns long tables (many rows, few columns) into large table (1 row, many columns)

Yes, <coe_single> can be gathered into <coe_list>, and can be a column in a <mom>.

Now you know how it works.

The best news is that, in most cases, and whether you have met retired Momocs before, you won’t have to bother with this2:

bot %>% efourier()

MomX and the tidyverse

MomX owes everything to the tidyverse and to the humans behind it. I deeply believe, it is not a trend or something, it is more the right way to do things in data science as a whole.

In particular, MomX :

  • is built on top of the tidyverse. For example, Momocs2 has no other dependencies outside those that the tidyverse also uses
  • embraces the tidy tools manifesto
  • is directly inspired, in its ecosystemic nature, by the tidyverse

MomX thus makes a massive use of dplyr, ggplot2, purrr and, on a lighter note, on magrittr pipe operators. They are a must and you haven’t heard of these, stop what you’re doing, close the door and go there :

https://r4ds.had.co.nz/

  1. Please drink responsibly↩︎

  2. I did it for you ;-)↩︎