Future of xsparql

[zacharywhitley]

@VladimirAlexiev @nunolopez It's nice to see some activity here. I've always thought that xsparql was a brilliant idea and have been disappointed that it hasn't gotten more interest from the community. I wanted to open this issue to see if we could have a discussion of why that is and what might move it forward.

[VladimirAlexiev]

I have the same sentiments...

In lieu of fixes, we should at least gather problems, resolutions and best practices.
Here's a start ("forked repository and attached jar" refers to https://github.com/patzomir/xsparql)

• Use "declare default element namespace" or always specify the namespaces of elements in xpath
• Unicode bug #2 writing to STDOUT is inferior compared to writing to -f #45: Unicode problems if writing to STDOUT.
  Workaround: use "-f file.ttl" to write to a file.
  Issue 45 shows the pipeline I use to add "base", prettify the output, and sort it by paragraph
• SQL ERROR: prefix cannot be null (JSE > 6): fixed in the forked repository and attached jar
• rewriter gets confused by attribute named "count" #7, rewriter gets confused by element name "group" #8:
  Certain names used in XPath expressions (eg group for an element or @count for an attribute) are mis-recognized as keywords.
  Workarounds:
  • parent/*[name()="group"]
  • parent/attribute::*[name()="count"]
• rewriter error messages are useless #9: The parser gives very unhelpful error messages, often showing an error location that is "upward" and "to the side" of the real problem.
  Mitigation: use the railroad grammar diagrams: https://rawgit2.com/VladimirAlexiev/xsparql/master/doc/grammar.xhtml
• If you don't use output "method=text", unnecessarily encodes XML special characters: see section below
• rewriter doesn't allow union xpaths #10, how to use an OR path (disjunction)? #39 Cannot use OR path (disjunction) or "union".
  Work around: select the two alternatives as a sequence of two elements
• grammar doesn't support embedded XML #14: Grammar doesn't support embedded XML.
  Workaround: use a WHERE VALUES lookup table (see below)
• URL guessing is conceptually wrong #15, Guessing for URI #16:
  Treats all strings starting with "http:" as URIs. This will lead to incorrect turtle file, when there are spaces or special characters in the text.
  Fixed in the forked repository and attached jar
• How to capture piece of XML to a literal as is? #38: How to capture piece of XML to a literal as is?
  This is needed for GeoSPARQL literals. Resolved by using an internal function "XSPARQL:_serialize"
• how to use WHERE VALUES for a lookup table? #41: how to use WHERE VALUES for a lookup table:
  It works, but is a bit fiddly since it requires a FILTER, and prints some "Prova.txt" crap on STDOUT
• how to preserve BASE? #42: how to specify "base"? It ignores "base".
  Workaround: use relative URLs freely, but post-process with "riot --syntax ttl --formatted ttl --base ...".
  This is a good idea anyway, to improve the formatting.
• undef is recognized, UNDEF is not #43: "UNDEF" not recognized.
  Workaround: use "undef" (lowercase)
• bind not allowed in where ??? #44: "bind" is not recognized in "where".

[nunolopes]

Hi, great to see activity and thanks for all the reports collected, maybe we should get all contributions into a central place?

Pinging @stefanbischof and @AxelPolleres.