dev-branch normalize.py reports errors badly

[holdenweb]

In an attempt to remove the meaningless and unhelpful traceback information when the notebook file is unreadable I now trap the exception and terminate, printing out the error string from the exception.

Unfortunately the only information I currently get is

NotJSONError('Notebook does not appear to be JSON: '{\n "metadata": {'',)

It would be much more helpful to provide the positional data about the point of JSON parsing failure, which I understand should be available (somewhere?)

[holdenweb]

I discovered that changing the parse_json() function in nbformat/current.py to read

def parse_json(s, **kwargs):
    """Parse a string into a (nbformat, dict) tuple."""
    try:
        d = json.loads(s, **kwargs)
    except ValueError as e:
        raise NotJSONError("Notebook does not appear to be JSON. Parser reports:\n   %s" % e)
    nbf = d.get('nbformat', 1)
    nbm = d.get('nbformat_minor', 0)
    return nbf, nbm, d

will report the error. Trunk on IPython's dev repository has changed that load and store code so comprehensively I don't think it's worth submitting an upstream patch.

[TheOpenBastion]

Though I wonder if it could be useful to have The Open Bastion as a
contributor to IPython, especially re these normalizing functions.

On Mon, Aug 18, 2014 at 2:34 PM, Steve Holden notifications@github.com
wrote:

I discovered that changing the parse_json() function in
nbformat/current.py to read

def parse_json(s, *_kwargs):
"""Parse a string into a (nbformat, dict) tuple."""
try:
d = json.loads(s, *_kwargs)
except ValueError as e:
raise NotJSONError("Notebook does not appear to be JSON. Parser reports:\n %s" % e)
nbf = d.get('nbformat', 1)
nbm = d.get('nbformat_minor', 0)
return nbf, nbm, d

will report the error. Trunk on IPython's dev repository has changed that
load and store code so comprehensively I don't think it's worth submitting
an upstream patch.

—
Reply to this email directly or view it on GitHub
#1 (comment)
.

Thank you,
Rachel Kelly
The Open Bastion
503 422 6171

[holdenweb]

I'm not sure how much the filtering alone will stabilize the issue, nor what improvements we might see from the upcoming ordering of the JSON dict keys. While we aren't doing anything revolutionary we probably should learn at least some parts of the code base and contribute what ideas we can.