Bizarro Pipe

“Bizarro Pipe” is a debugging tool for long dplyr pipelines. It is simply the idea that the glyph ->.; behaves like an eager replacement for the magrittr pipe %>% and makes single step debugging much easier.

library("babynames") # data package
library("dplyr")     # provides data manipulating functions.

library("magrittr")  # ceci n'est pas un pipe
library("ggplot2")   # for graphics

babynames %>%
    filter(name %>% substr(1, 3) %>% equals("Ste")) %>%
    group_by(year, sex) %>%
    summarize(total = sum(n)) %>%
    qplot(year, total, color = sex, data = ., geom = "line") %>%
    add(ggtitle('Names starting with "Ste"')) %>%
    print

To write the same pipeline using “Bizarro Pipe” we must make all the uses of first argument explicit with an extra “dot notation” in all of the “outer pipeline” (not actually a bad practice in terms of readability). By “outer pipes” we mean magrittr pipes that are in the outer pipeline and not inside the argument of an operator (i.e., not as with the “name %>% substr(1, 3) %>% equals("Ste")” inner pipeline).

babynames %>%
    filter(.,  name %>% substr(1, 3) %>% equals("Ste") ) %>%
    group_by(., year, sex) %>%
    summarize(., total = sum(n)) %>%
    qplot(year, total, color = sex, data = ., geom = "line") %>%
    add(., ggtitle('Names starting with "Ste"')) %>%
    print(.)

And then we convert all outer pipes (%>%) to Bizarro form (->.;) by a simple search and replace.

babynames ->.;
    filter(.,  name %>% substr(1, 3) %>% equals("Ste") ) ->.;
    group_by(., year, sex) ->.;
    summarize(., total = sum(n)) ->.;
    qplot(year, total, color = sex, data = ., geom = "line") ->.;
    add(., ggtitle('Names starting with "Ste"')) ->.;
    print(.)

This above form is now single step or per-line debuggable. We can step through each stage of the outer pipeline and examine the contents of “.” after each step (either by printing “.” or by using RStudio with RStudio -> Preferences -> General -> "Show .Last.value in environment listing" checked). We have a free video of the technique in action here.

It is a little disappointing we can’t replace the inner or sub-expression pipes with Bizarro Pipe; but they are part of an expression that is captured for row-wise execution in the filter and not something we could step through in the context of the outer pipeline anyway.

Also note: while we can think of ->.; as an eager operator we may need to force computation of a lazy structure (such as a tbl) by printing “.”.

Overall the trick is to train your eye to see ->.; as an atomic or indivisible glyph (with superhero logo).

The following image illustrates a pipeline (taken from here) without magrittr or dplyr (using only purrr verbs and Bizarro Pipe).

Bizarro Pipe

John Mount

2017-03-13