关于jsonlite包fromJSON函数JSON到R类型映射的疑问

Hey there! Let's unpack your questions about how jsonlite maps JSON structures to R data types—this is a common point of confusion, so let's break it down with your specific example.

1. Why are cars parsed as a data.frame and models as a list? #

Let's look at your JSON structure again:

{"cars": [ {"models" : [ {"type":"Fiesta"} ] } ] }

• The cars array contains objects with consistent structure (each object has exactly one key: models). By default, fromJSON() simplifies such arrays into data.frames, since each object can be treated as a row with columns matching the object's keys. That's why data$cars is a data.frame with one column named models.
• The models value in each cars object is itself a JSON array ([{"type":"Fiesta"}]). Since this array contains an object (not simple values like strings/numbers), and it's stored as a column in the cars data.frame, jsonlite can't simplify it into a regular vector (data.frame columns need to be vectors or lists). Instead, it keeps this as a list column—each entry in the models column is a list representing the inner JSON array.

2. What are the rules for mapping JSON entities to R types? #

jsonlite uses a set of heuristics (controlled mainly by the simplifyVector argument, which defaults to TRUE) to decide how to convert JSON to R types:

• JSON objects (key-value pairs):
  • If all values are of consistent length/structure (e.g., each value is an array of the same length, or a scalar), they're converted to a data.frame (columns = object keys, rows = values).
  • If values are mixed (e.g., some scalars, some arrays of different lengths), they're converted to a named list.
• JSON arrays:
  • If all elements are simple values (strings, numbers, booleans, NULL) of the same type, converted to an R vector.
  • If all elements are objects with identical key sets, converted to a data.frame (each object = row, keys = columns).
  • If elements are mixed types, or arrays (like your models array), converted to an R list.
• When simplifyVector = FALSE, all JSON structures are converted to nested lists with no simplification—this is useful if you want to preserve the exact JSON hierarchy.

3. Why does names(data$cars$models) return NULL? #

data$cars$models is a list where each element corresponds to the models array from a cars object. In your case, it's a list of length 1, and that single element is another list (list(type = "Fiesta")).

The reason names(data$cars$models) is NULL is because JSON arrays are ordered, unlabeled collections—they don't have keys for their elements. So when jsonlite parses the models JSON array into an R list, that list doesn't have names attached to its elements. If you want to access the type value, you'd use:

data$cars$models[[1]]$type

This drills into the first element of the models list, then accesses the named type entry inside that inner list.

内容的提问来源于stack exchange，提问作者Pioggia_91