FOTS: 'castable as' issues with numeric values

[Sorin Marian Nasoi]

Here are some examples of wrong results from 'castable as'.
ZORBA_WITH_BIG_INTEGER set to OFF/ON makes no difference:

-129 castable as xs:byte
returns true
http://www.w3.org/TR/xmlschema-2/#byte mentions the "minInclusive· to be -128" ...

1.7976931348623157E+308 castable as xs:decimal
returns true

2147483648 castable as xs:int
returns true

1.7976931348623157E+308 castable as xs:integer
returns true

-9223372036854775809 castable as xs:long
returns true for ZORBA_WITH_BIG_INTEGER set to ON, err:FOAR0002 otherwise which is correct

Basically the same error happens with every numeric data type.

More examples include (but are not limited to):
From the "prod-CastableExpr":

"cbcl-castable-byte-001,cbcl-castable-byte-002, cbcl-castable-decimal-007,cbcl-castable-decimal-008,cbcl-castable-decimal-009,cbcl-castable-decimal-010,cbcl-castable-duration-001,cbcl-castable-duration-002,cbcl-castable-int-001,cbcl-castable-int-002,cbcl-castable-integer-001,cbcl-castable-integer-002,cbcl-castable-integer-003,cbcl-castable-integer-004,cbcl-castable-long-001,cbcl-castable-long-002,cbcl-castable-negativeInteger-001,cbcl-castable-nonNegativeInteger-001,cbcl-castable-nonPositiveInteger-001,cbcl-castable-positiveInteger-001,cbcl-castable-short-001,cbcl-castable-short-002,cbcl-castable-unsignedByte-001,cbcl-castable-unsignedByte-002,cbcl-castable-unsignedInt-001,cbcl-castable-unsignedInt-002,cbcl-castable-unsignedLong-001,cbcl-castable-unsignedLong-002,cbcl-castable-unsignedShort-001,cbcl-castable-unsignedShort-002"

[Sorin Marian Nasoi (sorin.marian.nasoi)]

@Markos: please reassign if needed.
I assigned it to you because this could be related to some optimisation issue.

[Sorin Marian Nasoi (sorin.marian.nasoi)]

Reassigned the numeric failures to Paul as Markos suggested.

[Paul J. Lucas (paul-lucas)]

I figured out that the ATOMIC_CAST_MATRIX in root_typemanager.cpp was wrong for several types. According to "bzr blame," Markos is the author of this code. I fixed the obviously wrong matrix entries -- you're welcome. Assigning to Markos for review and follow-up.

[Markos Zaharioudakis (markos-za)]

Where is the fix in ATOMIC_CAST_MATRIX? I didn't see any commit in the trunk. Is it in a branch?
And no, I am not the original author of this code (Vinayak is).

[Sorin Marian Nasoi (sorin.marian.nasoi)]

Just wanted to mention that merge proposal
https://code.launchpad.net/~zorba-coders/zorba/bug-1077206/+merge/134797
was not a complete fix for this bug although linking the bug no. to a merge proposal in a way implies that the bug is fixed by the merge proposal.

for instance the query:
-129 castable as xs:byte
still returns true.

Same is true for the other queries mentioned in the original report.

[Markos Zaharioudakis (markos-za)]

Sorin, -129 castable as xs:byte returns false for me. Can you double check please?

[Markos Zaharioudakis (markos-za)]

With BIG_INTEGERS=ON, the following cases in the prod-CastableExpr fail still:

1. Castable-UnionType-1
2. Castable-ListType-2

This is not a bug, but a missing V3.0 extension: In V1.0, the target type must be an atomic type. In V3.0, it can be a simple type.

3. cbcl-castable-decimal-007
4. cbcl-castable-decimal-008
5. cbcl-castable-decimal-009
6. cbcl-castable-decimal-010
7. cbcl-castable-integer-001
8. cbcl-castable-integer-002
9. cbcl-castable-integer-003
10. cbcl-castable-integer-004

I don't understand why the expected result of these is "false". I think this is actually a FOTS bug (the expected result should be "true", assuming the implementation does not pose any upper bounds on integers and decimals)

[Sorin Marian Nasoi (sorin.marian.nasoi)]

Hi Markos: I can confirm that indeed after a clean build I see that 20 of the reported tests were fixed: thanks for the help.

Can you please file 2 bugs with W3C for the issue you reported:
- one for the missing XQuery 3.0 dependency in test cases Castable-UnionType-1 and Castable-ListType-2

- one for the other 8 test cases ( cbcl-castable-decimal-007 through cbcl-castable-decimal-010 and cbcl-castable-integer-001 through cbcl-castable-integer-004).
I have checked the latest version of the test set:
http://dev.w3.org/cvsweb/2011/QT3-test-suite/prod/CastableExpr.xml?rev=1.10;content-type=text%2Fplain
and there is no environment setting whatsoever for these test cases

In order to report a bug with W3C for FOTS please select these:
- first choose 'XPath / XQuery / XSLT'
- choose the 'XQuery 3 and XPath 3 Test Suite' component
HTH.

Adding the 2 bug reports with W3C will allow us to mark this bug as 'fix committed'.

[Sorin Marian Nasoi (sorin.marian.nasoi)]

Added lp:1090251 for the remaining unsolved test cases that were originally reported.

Marking this bug as "Fix committed".