How to handle json dump for inputs with CDATA?

[cehbrecht]

Description

When dumping (as json) a complex-input with format json then the data gets tagged with CDATA:

pywps/pywps/inout/inputs.py

Line 243 in 9fa56cc

data["data"] = u'<![CDATA[{}]]>'.format(out)

Is this necessary?

This needs to be handled when loading the json dump again ... currently this is not the case.

See also #444.

Environment

• operating system:
• Python version:
• PyWPS version: 4.2
• source/distribution
• git clone
• Debian
• PyPI
• zip/tar.gz
• other (please specify):
• web server
• Apache/mod_wsgi
• CGI
• other (please specify):

Steps to Reproduce

We have a workflow process with a workflow document in json:
https://github.com/roocs/rook/blob/858130631bf0a37c19a78e8e94961b7159846833/rook/processes/wps_orchestrate.py#L13

The json document is send with the wps request ... not as reference.

When we use the pywps scheduler extension the WPSRequest is dumped as a json file:

pywps/pywps/processing/job.py

Line 71 in 9fa56cc

def dump(self):

In this case the workflow document (json format) is tagged by CDATA:

pywps/pywps/inout/inputs.py

Line 243 in 9fa56cc

data["data"] = u'<![CDATA[{}]]>'.format(out)

When the json dump is loaded the CDATA tag is still part of the workflow document:

pywps/pywps/processing/job.py

Line 81 in 9fa56cc

def load(cls, filename):

... and the json loader for the workflow document will fail.
https://github.com/roocs/rook/blob/858130631bf0a37c19a78e8e94961b7159846833/rook/processes/wps_orchestrate.py#L52

Additional Information

PR #444

[cehbrecht]

@huard how would you handle this?

Should we add the json data as it is without CDATA?

pywps/pywps/inout/inputs.py

Line 227 in 9fa56cc

if self.data_format.mime_type in ["application/xml", "application/gml+xml", "text/xml"]:

Or when loading the json dump should the CDATA tag be removed?

pywps/pywps/inout/inputs.py

Line 217 in 9fa56cc

instance.data = json_input['data']

[huard]

Hi @cehbrecht ,

I remember banging my head about this, trying to find a solution that worked across the board. I think the issue is that the json content may contain characters that will confuse the xml parser, hence the need to put everything inside CDATA tags.

I thought that the XML parser removed the CDATA tags automatically. Sorry I don't have a better answer...

[cehbrecht]

fixed by #555