Datasets do not validate as XML when created with a string containing a specified encoding

[dalepotter]

Linked to #24, datasets with an encoding declared do not validate as XML.

This example shows the problem using code from the master branch (v0.3.0):

$ python
Python 3.6.0 (default, Dec 24 2016, 08:02:28) 
[GCC 4.2.1 Compatible Apple LLVM 7.0.2 (clang-700.1.81)] on darwin
Type "help", "copyright", "credits" or "license" for more information.
>>> import iati
>>> dataset_xml_declaration_no_encoding = iati.Dataset("""
... <?xml version="1.0"?>
... <iati-activities version="xx">
...   <iati-activity>
...     <iati-identifier></iati-identifier>
...     <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...     <title>
...       <narrative>Xxxxxxx</narrative>
...     </title>
...     <description>
...       <narrative>Xxxxxxx</narrative>
...     </description>
...     <participating-org role="xx"></participating-org>
...     <activity-status code="xx"/>
...     <activity-date type="xx" iso-date="2013-11-27"/>
...     <activity-date type="xx" iso-date="2013-11-27">
...       <narrative>Xxxxxxx</narrative>
...     </activity-date>
...   </iati-activity>
... </iati-activities>
... """)
>>> iati.validator.is_xml(dataset_xml_declaration_no_encoding)
True

vs. the same dataset with and encoding="UTF-8"? declared:

>>> dataset_xml_declaration_with_encoding = iati.Dataset("""
... <?xml version="1.0" encoding="UTF-8"?>
... <iati-activities version="xx">
...   <iati-activity>
...     <iati-identifier></iati-identifier>
...     <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...     <title>
...       <narrative>Xxxxxxx</narrative>
...     </title>
...     <description>
...       <narrative>Xxxxxxx</narrative>
...     </description>
...     <participating-org role="xx"></participating-org>
...     <activity-status code="xx"/>
...     <activity-date type="xx" iso-date="2013-11-27"/>
...     <activity-date type="xx" iso-date="2013-11-27">
...       <narrative>Xxxxxxx</narrative>
...     </activity-date>
...   </iati-activity>
... </iati-activities>
... """)
>>> iati.validator.is_xml(dataset_xml_declaration_with_encoding)
False

This latter XML (pastebin link for convenience) does validate as XML using two online XML validation sites: codebeautify and truugo

[hayfield]

Noting that relevant tests are in test_data.py#TestDatasetWithEncoding

[hayfield]

An XML string must not have any leading whitespace, as both these examples do.

https://github.com/IATI/pyIATI/blob/59588100911f4fd17a4012c0c2bc9632cc20efbd/iati/data.py#L87 undertakes some amount of stripping of leading and trailing whitespace, though an explicit encoding may cause complications. There is currently only one test relating to leading whitespace - this doesn't seem fully comprehensive!

https://github.com/IATI/pyIATI/blob/59588100911f4fd17a4012c0c2bc9632cc20efbd/iati/tests/test_data.py#L46-L53

[hayfield]

>>> dataset_xml_declaration_with_encoding_2 = iati.Dataset("""<?xml version="1.0"?>
...  <iati-activities version="xx">
...    <iati-activity>
...      <iati-identifier></iati-identifier>
...      <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...      <title>
...        <narrative>Xxxxxxx</narrative>
...      </title>
...      <description>
...        <narrative>Xxxxxxx</narrative>
...      </description>
...      <participating-org role="xx"></participating-org>
...      <activity-status code="xx"/>
...      <activity-date type="xx" iso-date="2013-11-27"/>
...      <activity-date type="xx" iso-date="2013-11-27">
...        <narrative>Xxxxxxx</narrative>
...      </activity-date>
...    </iati-activity>
...  </iati-activities>
...  """)
>>> iati.validator.is_xml(dataset_xml_declaration_with_encoding_2)
True

As premised, this is a problem that the provided string is not valid XML because it contains leading whitespace. This is therefore a problem with an explicit encoding in combination with leading whitespace (the automatic removal of which is deemed to be a feature of pyIATI).

I will update the title to better reflect this.

[dalepotter]

I think the wrong string was tested! With no leading whitespace the same results come back...

No whitespace and no encoding

>>> dataset_xml_declaration_with_encoding_2 = iati.Dataset("""<?xml version="1.0"?>
...   <iati-activities version="xx">
...     <iati-activity>
...       <iati-identifier></iati-identifier>
...       <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...       <title>
...         <narrative>Xxxxxxx</narrative>
...       </title>
...       <description>
...         <narrative>Xxxxxxx</narrative>
...       </description>
...       <participating-org role="xx"></participating-org>
...       <activity-status code="xx"/>
...       <activity-date type="xx" iso-date="2013-11-27"/>
...       <activity-date type="xx" iso-date="2013-11-27">
...         <narrative>Xxxxxxx</narrative>
...       </activity-date>
...     </iati-activity>
...   </iati-activities>
...   """)
>>> iati.validator.is_xml(dataset_xml_declaration_with_encoding_2)
True

No whitespace and a UTF-8 encoding

>>> dataset_xml_declaration_with_encoding_3 = iati.Dataset("""<?xml version="1.0" encoding="UTF-8"?>
...   <iati-activities version="xx">
...     <iati-activity>
...       <iati-identifier></iati-identifier>
...       <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...       <title>
...         <narrative>Xxxxxxx</narrative>
...       </title>
...       <description>
...         <narrative>Xxxxxxx</narrative>
...       </description>
...       <participating-org role="xx"></participating-org>
...       <activity-status code="xx"/>
...       <activity-date type="xx" iso-date="2013-11-27"/>
...       <activity-date type="xx" iso-date="2013-11-27">
...         <narrative>Xxxxxxx</narrative>
...       </activity-date>
...     </iati-activity>
...   </iati-activities>
...   """)
>>> iati.validator.is_xml(dataset_xml_declaration_with_encoding_3)
False

The error log tells us more...

>>> err_log = iati.validator.validate_is_xml(dataset_xml_declaration_with_encoding_3)
>>> len(err_log)
1
>>> err_log[0].name
'err-not-xml-not-string'
>>> err_log[0].info
"The value provided is a `<class 'str'>` rather than a `str`."

But... A workaround?!

However, when it is encoded to a bytes object all is well...

>>> dataset_xml_declaration_with_encoding_3 = iati.Dataset("""<?xml version="1.0" encoding="UTF-8"?>
...   <iati-activities version="xx">
...     <iati-activity>
...       <iati-identifier></iati-identifier>
...       <reporting-org type="xx" ref="xx"><narrative>Organisation name</narrative></reporting-org>
...       <title>
...         <narrative>Xxxxxxx</narrative>
...       </title>
...       <description>
...         <narrative>Xxxxxxx</narrative>
...       </description>
...       <participating-org role="xx"></participating-org>
...       <activity-status code="xx"/>
...       <activity-date type="xx" iso-date="2013-11-27"/>
...       <activity-date type="xx" iso-date="2013-11-27">
...         <narrative>Xxxxxxx</narrative>
...       </activity-date>
...     </iati-activity>
...   </iati-activities>
...   """.encode())
>>> iati.validator.is_xml(dataset_xml_declaration_with_encoding_3)
True

@hayfield mentioned that all tests for validation use bytes objects - I'd suggest adding some tests where we test strings.

[hayfield]

Due to the re-ordering of Dataset-creation operations in #286, the error occurs earlier under that branch. As such, that may be a better place to start from (also because it's a change that looks to explicitly separate how bytes and str objects are treated).

[hayfield]

The underlying error raised by lxml is: ValueError('Unicode strings with encoding declaration are not supported. Please use bytes input or XML fragments without declaration.',) - will look to improve the visibility of this message.

[hayfield]

Changing from bug to enhancement since lxml does not support this feature, and so this would be some additional pyIATI functionality to convert strs to bytes where required.

[hayfield]

NOTE: This is only a problem at Python 3 due to the changes to what a str is.