将 JSON 数组读入类似 Julia DataFrame 的类型 [英] Reading JSON array into Julia DataFrame-like type

问题描述

给定一个 JSON 文件，JSON 包会愉快地解析它.但是，如果我希望它作为 DataFrame (或任何其他列式数据结构)，那么获得它的好方法是什么?

Given a JSON file, the JSON package happily parses it. But if I would like it as a DataFrame (or any other columnar data structure), what would be a good way to get it?

目前，例如，我有:

using JSON
using DataFrames

json_str = """
[{ "color": "red", "value": "#f00" }, { "color": "green", "value": "#0f0" }, 
 { "color": "blue", "value": "#00f" }, { "color": "cyan", "value": "#0ff" }, 
 { "color": "magenta", "value": "#f0f" }, { "color": "yellow", "value": "#ff0" }, 
 { "color": "black", "value": "#000" } ]
  """

function jsontodf(a)
    ka = union([keys(r) for r in a]...)
    df = DataFrame(;Dict(Symbol(k)=>get.(a,k,NA) for k in ka)...)
    return df
end

a = JSON.Parser.parse(json_str)
jsontodf(a)

导致:

7×2 DataFrames.DataFrame
│ Row │ color     │ value  │
├─────┼───────────┼────────┤
│ 1   │ "red"     │ "#f00" │
│ 2   │ "green"   │ "#0f0" │
│ 3   │ "blue"    │ "#00f" │
│ 4   │ "cyan"    │ "#0ff" │
│ 5   │ "magenta" │ "#f0f" │
│ 6   │ "yellow"  │ "#ff0" │
│ 7   │ "black"   │ "#000" │

并且还处理一些带有 NA 的缺失字段.有什么更清洁/更快的(Julia v0.6+)吗?

and also handles some missing fields with NAs. Anything cleaner / faster (Julia v0.6+) ?

推荐答案

我已经挖出了这个老问题，现在我们有了更好的解决方案，从 DataFrames.jl 0.18.0 开始.

I have dug out this old question, and now we have a better solution for it as of DataFrames.jl 0.18.0.

如果 JSON 中的所有条目都具有相同的字段，您可以编写:

If all entries in JSON have the same fields you can write:

reduce(vcat, DataFrame.(a))

如果您必须处理每个字典中不同字段的可能性，请编写:

If you have to handle the possibility of different fields in each dict then write:

vcat(DataFrame.(a)..., cols=:union)

如果 a 有很多条目，这可能会有点问题.我刚刚提交了一个 PR，这样你也可以写:

This can be slightly problematic if a has a lot of entries as it does splatting. I have just submitted a PR so that you will be also able to write:

reduce(vcat, DataFrame.(a), cols=:union)

在不久的将来.

这篇关于将 JSON 数组读入类似 Julia DataFrame 的类型的文章就介绍到这了，希望我们推荐的答案对大家有所帮助，也希望大家多多支持IT屋！