.discover fails on whitespace

[dgtlmoon]

Other project https://github.com/fcurella/python-datauri/ like has no problems (pip3 install python-datauri)

FAIL
<img src="data:image/jpeg;base64, iVBORw0KGgoAAAANSUhEUgAgglpyJYUQnIAABCJJggg==">

PASS
<img src="data:image/jpeg;base64,iVBORw0KGgoAAAANSUhEUgAgglpyJYUQnIAABCJJggg==">

[wbolster]

hi, thanks for your contribution.

according to rfc 2397 the syntax is:

       dataurl    := "data:" [ mediatype ] [ ";base64" ] "," data
       mediatype  := [ type "/" subtype ] *( ";" parameter )
       data       := *urlchar
       parameter  := attribute "=" value

...which means that whitespace is not allowed.

the data: uri page on wikipedia contains:

In this example, the lines are broken for formatting purposes. In actual URIs, including data URIs, control characters (ASCII 0 to 31, and 127) and spaces (ASCII 32) are "excluded characters". This means that whitespace characters are not permitted in data URIs.

though it continues like this:

However, in the context of HTML 4 and HTML 5, linefeeds within an element attribute value (such as the "src" above) are ignored[citation needed]. So the data URI above would be processed ignoring the linefeeds, giving the correct result. But note that this is an HTML feature, not a data URI feature, and in other contexts, it is not possible to rely on whitespace within the URI being ignored.

i am not sure what the best behaviour would be here.

(note: the behaviour of another library is not leading. that other library also seems to reparse on every attribute access, which this library will never do.)

[ghost]

It seems to me like it should be safe to strip leading/trailing whitespace if the data is base64-encoded. I would be in support of doing that for convenience's sake. Other data types may care about whitespace.

[wbolster]

from #8 and #8 (comment)

i personally think code should either be strict (no whitespace, per spec) or loose (strip all whitespace, useful for html-ish data), not something in between.