add indent support to `to_json` method

[phretor]

Sometimes you just want to have a quick look at your data in IPython and having to_json to support indent (as both json.dumps and simplejson.dumps do) would be very handy.

A workaround, which however doesn't always work, is to do json.dumps(series.to_dict()), which unfortunately is annoying when you don't have a JSON-serializable type (e.g., numpy.bool_).

I'm leaving this here as a food for discussion.

[ludaavics]

Another workaround: df_json_pretty = json.dumps(json.loads(df.to_json()), indent=2)

[devforfu]

Is it still on the list to be implemented?

[TomAugspurger]

Still open. Feel free to work on it if you want, but the current json isn't the friendliest.

[devforfu]

@TomAugspurger Ok, got it! I'll try to figure out if there is a straightforward way to deal with it.

[kylebarron]

It looks like the JSON writer uses Pandas' own JSON writer:

pandas/pandas/io/json/json.py

Line 7 in 011b79f

import pandas._libs.json as json

It looks like eventually it calls objToJSON in C here:

pandas/pandas/_libs/src/ujson/python/objToJSON.c

Lines 2317 to 2321 in 011b79f

	PyObject *objToJSON(PyObject *self, PyObject *args, PyObject *kwargs) {
	static char *kwlist[] = {
	"obj", "ensure_ascii", "double_precision", "encode_html_chars",
	"orient", "date_unit", "iso_dates", "default_handler",
	NULL};

So if I'm not mistaken, to add an indent option, you'd need to change the C code?

[TomAugspurger]

Yep that sounds right. FYI, I think ripping out that JSON code and replacing it with something more maintainable (maybe based on Apache Arrow's JSON support) is on the medium to long-term agenda.

…

On Mon, Nov 12, 2018 at 9:28 AM Kyle Barron ***@***.***> wrote: It looks like the JSON writer <https://github.com/pandas-dev/pandas/blob/011b79fbf73b45313b47c08b4be1fc07dcb99365/pandas/io/json/json.py#L107-L117> uses Pandas' own JSON writer: https://github.com/pandas-dev/pandas/blob/011b79fbf73b45313b47c08b4be1fc07dcb99365/pandas/io/json/json.py#L7 It looks like eventually it calls objToJSON in C here: https://github.com/pandas-dev/pandas/blob/011b79fbf73b45313b47c08b4be1fc07dcb99365/pandas/_libs/src/ujson/python/objToJSON.c#L2317-L2321 So if I'm not mistaken, to add an indent option, you'd need to change the C code? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#12004 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIiHGfDOPOU98wggIOgkOVEaLVrPgks5uuZOKgaJpZM4HBmld> .

[kylebarron]

Is the current JSON writer believed to be much faster than the Python standard library's JSON writer?

[TomAugspurger]

Yep. It also handles types not supported by the stdlib's json writer (notably numpy types).

…

On Mon, Nov 12, 2018 at 9:55 AM Kyle Barron ***@***.***> wrote: Is the current JSON writer believed to be much faster than the Python standard library's JSON writer? — You are receiving this because you were mentioned. Reply to this email directly, view it on GitHub <#12004 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/ABQHIvmwtHI6ob3U9Tb4WZVI97EsrOd8ks5uuZn9gaJpZM4HBmld> .

[botenvouwer]

I agree on this, JSON is also meant to be human readable. Right now this is just annoying.