PandasJSON.jl

A Julia package for reading and writing Pandas dataframes in JSON format.

pkg> add PandasJSON

First we load relevant packages. In this tutorial we use DataFrames.jl for our tabular data, but any Tables.jl-compatible data structure will do.

julia> using PandasJSON, DataFrames

Now we create a table with two columns and some missing data.

julia> df = DataFrame(x=[1,2,3], y=[true,false,missing])
3×2 DataFrame
 Row │ x      y
     │ Int64  Bool?
─────┼────────────────
   1 │     1     true
   2 │     2    false
   3 │     3  missing

Now let's save this to disk in JSON format and take a look at the resulting file.

julia> PandasJSON.write("example.json", df)

julia> println(read("example.json", String))
{"y":{"1":false,"0":true,"2":null},"x":{"1":2,"0":1,"2":3}}

Finally we read the JSON file back as a table and convert it to a DataFrame. We could do the same thing with a JSON file written in Python by Pandas.

julia> df = PandasJSON.read("example.json", DataFrame)
3×2 DataFrame
 Row │ x      y
     │ Int64  Bool?
─────┼────────────────
   1 │     1     true
   2 │     2    false
   3 │     3  missing

Note: These functions have an optional orient keyword argument, which controls how the tabular data is represented as a JSON structure. The default in both Pandas and PandasJSON is orient="split", so with default parameters everything should be compatible.

You should use this argument if either:

• You are reading data which set the orient to something non-default.
• You would like to guarantee row and column ordering is correct (split, table or values) or require more column type information to be stored (table).

If you are not sure, you can use guess_orient:

julia> PandasJSON.write("example.json", df, orient="table")

julia> PandasJSON.guess_orient("example.json")
1-element Vector{Symbol}:
 "table"

julia> df = PandasJSON.read("example.json", DataFrame, orient="table")
3×2 DataFrame
 Row │ x      y
     │ Int64  Bool?
─────┼────────────────
   1 │     1     true
   2 │     2    false
   3 │     3  missing

Read the docstrings for more details and keyword arguments.

• PandasJSON.read(file, [type]): Read a JSON file as a table.
• PandasJSON.write(file, table): Write a table to the file in JSON format.
• PandasJSON.guess_orient(file): Guess the orient parameter used to write the given file.

Note that PandasJSON.read should behave identically to pandas.read_json and PandasJSON.write should behave identically to pandas.DataFrame.to_json, including the behaviour of any supported keyword arguments. Any deviation is considered a bug.

There are many other Julia packages for reading and writing tabular data in formats supported by Pandas. We recommend using one of these instead if possible.