I'm trying to convert a Polars dataframe to a JSON object, but I can't seem to find a way to change the format of it between row/col orientation. In Pandas, by default, it creates a column-oriented JSON object, and it can be changed easily, but in Polars it's row-oriented and I can't seem to find a way to change that, since the write_json method doesn't actually receives any argument other than the target filename if you want to save it directly.
Any idea how to achieve that?
It is possible to achieve all orientations in Polars as well. It just requires a bit more creativity and json.dumps() instead of write_json alone. I've compiled a mapping from Pandas different orientations to Polars equivalent below. Using the same dataframe for each example.
import polars as pl
df = pl.DataFrame({
"A": [1, 2, 3],
"B": [4, 5, 6]
})
Index one seems to be a bit overcomplicated than needed, maybe there is a better way of doing that?
| Orientation | Pandas | Polars |
|---|---|---|
| Records | orient="records" |
df.write_json() |
| Columns | orient="columns" |
json.dumps(df.to_dict(as_series=False)) |
| Index | orient="index" |
json.dumps({i: row for i, row in enumerate(df.iter_rows(named=True))}) |
| Values | orient="values" |
json.dumps(df.to_numpy().tolist()) |
To achieve a column-oriented JSON structure, You need to transform the DataFrame to a Dictionary before exporting Example:
import polars as pl
import json
df = pl.DataFrame({"A": [1, 2, 3], "B": [4, 5, 6]})
# json_data = df.write_json() # I guess you have tried this
# Convert to column-oriented JSON
# json_data = json.dumps(df.to_dict(as_series=False), indent=4)
# Convert to column-oriented JSON without newlines
json_data = json.dumps(df.to_dict(as_series=False), separators=(',', ':'))
print(json_data)
Output:
{"A": [1, 2, 3], "B": [4, 5, 6]}
indent=4 then you'll get JSON without newlines.