blog.md

title	author	date	categories	tags	slug	lastmod	comment	toc	autoCollapseToc	postMetaInFooter	hiddenFromHomePage	contentCopyright	reward	mathjax	mathjaxEnableSingleDollar	mathjaxEnableAutoNumber	hideHeaderAndFooter	flowchartDiagrams	sequenceDiagrams
Iterating over the lines of a data.frame with purrr	rstats	2020-11-21	tidyverse	purrr	iterating-over-the-lines-of-a-data-frame-with-purrr	2020-11-21T17:41:43+02:00	false	true	false	false	false	false	false	false	false	false	false	enable options false	enable options false

I sometimes have a function which takes some parameters and returns a data.frame as a result. Then I have a data.frame where each row of it is a set of parameters. So I like to apply the function to each row of the parameter-data.frame and rbind the resulting data.frames.

There are several ways to do it. Let's have a look:

The function ...

So let's build a simple function we can use

https://gist.github.com/77b37ed4f28a03460b2372a1afba1573

https://gist.github.com/e600e6851e02b45538091f12cc1c5883

So this function takes three arguments and returns a data.frame. The length of the data.frame depends on the last parameter.

... and its parameters

So now we have several tuples of paramters. Each tuple is a row of our parameter-data.frame:

https://gist.github.com/83ca77d693bcbb46fff66621970f44d7

https://gist.github.com/c5fa0c41ec9ebc639d6a3e0af15aac15

So now we want to apply our function three times, one time for each row of the data.frame parameters.

Iterating with ...

There are several ways to interate.

... a for-loop

The most common way in programming is a for-loop:

https://gist.github.com/57c663ea0bfd59c7b4da41dcdfb2ea51

https://gist.github.com/bdcfd7f53a77f19c90d00ce20df8a96b

That's very ugly: You have to initialize the result-data.frame and it's slow. Whenever you want to use a for-loop in R step back and think about using something else.

... lapply()

Instead of for-loops you should use apply or one of its derivates. But apply works with lists. data.frames are lists but column-wise ones.

So we need to split the data.frame parameters into a list rowwise using split. Then we can apply my_function. Then we use do.call(rbind, x) do merge the results into one data.frame.

https://gist.github.com/882f0cf309d896345315d550325ec00f

https://gist.github.com/f97e03955b1c14f2e5c10bc3c23cde30

That's a lot more R-like. But the winner is:

... pmap_dfr() out of the purrr-package

The most elegant way I know of is purr's pmap_dfr

https://gist.github.com/2534bde812e39ea26f1f7f3f0193e5d4

https://gist.github.com/5ce3211a9d0c4f2ec1be8753671d4fae

pmap_dfr respects the column-names and parameter-names of the function. So you can mix them in the parameter-data.frame:

https://gist.github.com/edce95cae72ed09b0933509456a3a473

https://gist.github.com/957784a39f2e176ef22f904d1f1287ce