parseQuery decode the query string which cause invalid URI

[kfrawee]

I have the following ttl file:

@prefix : <https://www.example.co/reserved/language#> .

<https://www.example.co/reserved/root> :_id "01G39WKRH76BGY5D3SKDHJP2SX" ;
    :transcript%20data [ :_id "01G39WKRH7JYRX78X7FG4RCNYF" ;
            :_key "transcript%20data" ;
            :value "value" ;
            :value_id "01G39WKRH7PVK1DXQHWT08DZA8" ] .

And I have the following query:

q = """
PREFIX : <https://www.example.co/reserved/language#>

    SELECT  ?o 
    WHERE { ?s :transcript%20data/:value ?o . }
""" 

While trying to query the graph I got from the ttl file I got the following error:
https://www.example.co/reserved/language#transcript data does not look like a valid URI, trying to serialize this will break.
As you see, parseQuery has decoded the "%20" to a space " " which cases invalid URI. And this will return False while passed to _is_valid_uri function.

I've tested the query on different SPARQL engines and it is valid and works as expected.
So, what do you advise? to make the query valid and get the required results?

I am using rdflib Version: 6.1.1 on macOS Monterey 12.4

[kfrawee]

@aucampia @gjhiggins

[aucampia]

@kfrawee will look when I have time

[ghost]

I have the following ttl file:
...
So, what do you advise? to make the query valid and get the required results?

Looks like an issue with the SPARQL parser so it might be some time before it's fixed. In the interim, you can work around the problem by replacing %20 with _

[ghost]

fwiw:

1. I think the causal regexp is

   rdflib/rdflib/plugins/sparql/parser.py

   Line 221 in 6ed2ef4

   PN_LOCAL = Regex(

2. dunno if it's of any practical use in this instance but it may be worth taking note of / comparing with the example SPARQL 1.1 parser in the pyastbuilder repos

[ghost]

Ah, I was close but not close enough.

I'm still exploring the ramifications but if you're prepared to make a small edit to

rdflib/rdflib/plugins/sparql/parser.py

Line 230 in 6ed2ef4

def _hexExpand(match):

there is a temporary workaround that would enable you to work with the original data:

 def _hexExpand(match):
-    return chr(int(match.group(0)[1:], 16))
+    if match.group(0) == "%20":
+        return match.group(0)
+    else:
+        return chr(int(match.group(0)[1:], 16))

[kfrawee]

@gjhiggins

if match.group(0) == "%20":

It is not specific for "%20". I am using urllib.parse.quote to URL encode the ttl and quires. So, the same behavior happens to the rest of the characters.

[ghost]

It is not specific for "%20" ... the same behavior happens to the rest of the characters.

Yes, indeed so. I addressed the entire set of percent encoded reserved characters in the PR #1959 I submitted.