map_df Explained

map_df.rmd

---
title: "map_df Example"
author: "Alex Gold"
date: "1/29/2020"
output: html_document
---

```{r setup, include=FALSE}
library(tidyverse)
```

From the docs for map_df, the function signature of map_df is `map_df(.x, .f, ...)`. `.f` can be a function or character, if a character, it's used to create an extractor function. If it's a character, anything in `...` gets passed on to the extractor function.

By providing `my_data %>% map_df("attributes", "features")`, `.f = "attributes"`, and `...  = "features"`. 

Guessing at what your data looks like since you said, `my_data[["features"]][["attributes]]` works...
```{r}
(my_data <- list(
  features = list(
    attributes = tibble(
      dat = 1:3, 
      dat2 = 4:6
    ), 
    other_thing = 1:4
  ), 
  other_thing2 = 1:3
)
)
```

So the df I want is in

```{r}
my_data[["features"]][["attributes"]]
```

This doesn't work
```{r}
my_data %>% map_df("features", "attributes")
```

because `map_df` is mapping through the first nested layer, so it's _inside_ the features list when it's trying to do the first extraction. In other words, the function is first looking at `my_data$features` and trying to extract `"features"`, but there isn't an object named `"features"` inside `my_data$features`.

There is an object named `"attributes"`, so this works
```{r}
my_data %>% map_df("attributes")
```

This does too
```{r}
my_data %>% map_df("attributes", "features")
```
Because the argument `"features"` is being passed as an additional argument to the extractor function and not actually doing anything.

It's kinda dangerous to do it this way -- `map` is going across all of the sub-lists in `my_data`, so it's just lucky that there are no other `features` or `attributes` for it to give back things you _don't_ want.

In this case, I don't need a `map` function at all because I'm not actually iterating over anything. The most succint way to do this in a pipe is
```{r}
my_data %>% magrittr::extract2(c("features", "attributes"))
```