JSON - Parse mixed types as string in JSON reader

[karthikeyann]

Description

Addresses #14239

This PR adds an option to read mixed types as string columns.
It also adds related functional changes to nested JSON reader (libcudf, cuDF-python, Java).

Details:

• Added new option mixed_types_as_string bool in json_reader_options
• This feature requires 2 things: finding end of struct/list nodes, parse struct/list type as string.
• For Struct and List, node_range_end was node_range_begin+1 earlier (since it was not used anywhere). Now it is calculated properly by copying only struct and list tokens and their node_range_end is calculated. (Since end token is child of begin token, scattering end token's index to parent' token's corresponding node's node_range_end will get the node_range_end of List and Struct nodes).
• In reduce_to_column_tree() (which infers the schema), the list and struct node_range_end are changed to node_begin+1 so that it does not copy entire list/struct strings to host for column names.
• reinitialize_as_string reinitializes an initialized column as string.
• Mixed type columns are parsed as strings since their column category is changed to NC_STR.
• Added tests

Checklist

• I am familiar with the Contributing Guidelines.
• New or existing tests cover these changes.
• The documentation is up to date with these changes.

[karthikeyann]

I keep getting the following error sometimes while reading in python, and it crashes. (but libcudf read_json in c++ unit tests works).

Traceback (most recent call last):
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/asyncio/selector_events.py", line 256, in _add_reader
    key = self._selector.get_key(fd)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/selectors.py", line 193, in get_key
    raise KeyError("{!r} is not registered".format(fileobj)) from None
KeyError: '0 is not registered'

During handling of the above exception, another exception occurred:

Traceback (most recent call last):
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/bin/ipython", line 10, in <module>
    sys.exit(start_ipython())
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/IPython/__init__.py", line 130, in start_ipython
    return launch_new_instance(argv=argv, **kwargs)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/traitlets/config/application.py", line 1075, in launch_instance
    app.start()
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/IPython/terminal/ipapp.py", line 317, in start
    self.shell.mainloop()
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/IPython/terminal/interactiveshell.py", line 911, in mainloop
    self.interact()
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/IPython/terminal/interactiveshell.py", line 896, in interact
    code = self.prompt_for_code()
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/IPython/terminal/interactiveshell.py", line 839, in prompt_for_code
    text = self.pt_app.prompt(
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/prompt_toolkit/shortcuts/prompt.py", line 1026, in prompt
    return self.app.run(
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/prompt_toolkit/application/application.py", line 1021, in run
    return asyncio.run(coro)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/asyncio/runners.py", line 44, in run
    return loop.run_until_complete(main)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/asyncio/base_events.py", line 649, in run_until_complete
    return future.result()
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/prompt_toolkit/application/application.py", line 905, in run_async
    return await _run_async(f)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/prompt_toolkit/application/application.py", line 750, in _run_async
    with self.input.raw_mode(), self.input.attach(
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/contextlib.py", line 135, in __enter__
    return next(self.gen)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/site-packages/prompt_toolkit/input/vt100.py", line 165, in _attached_input
    loop.add_reader(fd, callback_wrapper)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/asyncio/selector_events.py", line 331, in add_reader
    self._add_reader(fd, callback, *args)
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/asyncio/selector_events.py", line 258, in _add_reader
    self._selector.register(fd, selectors.EVENT_READ,
  File "/home/karthikeyan/dev/rapids/compose/etc/conda/cuda_12.0/envs/rapids/lib/python3.10/selectors.py", line 360, in register
    self._selector.register(key.fd, poller_events)
OSError: [Errno 22] Invalid argument

If you suspect this is an IPython 8.20.0 bug, please report it at:
    https://github.com/ipython/ipython/issues
or send an email to the mailing list at [email protected]

You can print a more detailed traceback right now with "%tb", or use "%debug"
to interactively debug it.

Extra-detailed tracebacks for bug-reporting purposes can be enabled via:
    %config Application.verbose_crash=True

[andygrove]

LGTM. Thanks @karthikeyann

[jlowe]

Same comments as @andygrove otherwise lgtm from Java perspective.

[elstehle]

Flushing comments on the changes to json_tree.cu. Just a few minor comments/questions.

Nice work! Changes to identify the list/struct ends are shorter than what I had expected, mostly because you managed to reuse the existing utilities where possible 👍

[elstehle]

Approving for cpp changes.
Did you get a chance to check the performance numbers?

[bdice]

Comments attached. Overall I'm very pleased that this is < 500 lines and the implementation looks pretty good!

[bdice]

I left a couple small comments on naming, otherwise approving.

[shrshi]

How about renaming the function as propagate_parent_to_children_except_first? It would made the nodes to which the parent id is copied clearer from the function name.

[karthikeyann]

It propagates from first siblings to other siblings, provided first sibling already has some data.
I renamed it to propagate_first_sibling_to_other.

[shrshi]

Thank you for the clarifications! Approved.

[karthikeyann]

/merge