Opened 7 years ago
Closed 6 years ago
#1689 closed enhancement (worksforme)
Systemtest_rename rasql's test cases homogeneously
Reported by: | Bang Pham Huu | Owned by: | Dimitar Misev |
---|---|---|---|
Priority: | minor | Milestone: | Future |
Component: | systemtest | Version: | development |
Keywords: | Cc: | Dimitar Misev | |
Complexity: | Medium |
Description (last modified by )
In systemtest/testcases_mandatory (e.g: http://rasdaman.org/browser/systemtest/testcases_mandatory/test_select/queries)
the test queries are not homogeneous with systemtest/testcases_services (e.g: http://rasdaman.org/browser/systemtest/testcases_services/test_wcps/queries)
They should have:
+ Test number prefix (e.g: 001-arcosbig.rasql).
+ Test errors needs .error.rasql to recognize (current I need to check oracle file to know this test is error, not only in my system).
@Dragi: The ticket is about checking rasdaman/systemtest with any queries
folders and add the prefix with increasing numbers if they don't exist, e.g:
systemtest/testcases_mandatory/test_clipping/queries/cline_bothvertout.rasql -> systemtest/testcases_mandatory/test_clipping/queries/001-cline_bothvertout.rasql
NOTE:
if you change test files in queries folder (e.g: test_clipping/), check if test files also listed in test_clipping/known_fails and rename them accordingly.
and this is more frustrating, check folder oracles of test cases without prefix numbers and see which oracles are expected error cases. Add ".error." before the file extension for the corresponding query in queries folder and oracle file in oracle folder, e.g:
Open this file, it is an oracle with expected error from Rasql query, but there is no way to tell it is working query or non working query from its name:
systemtest/testcases_mandatory/test_clipping/oracle/e_clinestr_rg_card1.oracle rename to (note the prefix increasing number) systemtest/testcases_mandatory/test_clipping/oracle/001-e_clinestr_rg_card1.error.oracle and systemtest/testcases_mandatory/test_clipping/queries/001-e_clinestr_rg_card1.error.rasql
Change History (7)
comment:1 by , 7 years ago
Milestone: | 9.5 → Future |
---|
comment:2 by , 6 years ago
Cc: | added |
---|---|
Description: | modified (diff) |
Milestone: | Future → 9.7 |
Owner: | changed from | to
Status: | new → assigned |
comment:3 by , 6 years ago
Description: | modified (diff) |
---|
comment:4 by , 6 years ago
Milestone: | 9.7 → Future |
---|---|
Owner: | changed from | to
Priority: | major → minor |
comment:6 by , 6 years ago
The main benefit is I just need the preifix number when one test is failed, I can open test query or test oraclr or test output quickly.
It is very inconvenient when searching a test with more or less same names.
It is best to make them with patterns and a student can do it.
comment:7 by , 6 years ago
Resolution: | → worksforme |
---|---|
Status: | assigned → closed |
I'm not convinced by your reason, will close this ticket.
I'm not sure if this is a good idea anymore honestly. Without the number prefix it's possible to group queries by prefix (e.g. a bunch of different tests for arccos-big.rasql, arccos-small.rasql, etc). With the numbers we lose this grouping and get something like 001-arccos-big.rasql, 323-arccos-small.rasql, …
What's the advantage of this number prefix? Order of query evaluation is fixed. Although this can be emulated easily within groups where it could matter, e.g.
Not sure why is this important? If the test passes it's all fine, irrelevant of whether the oracle contains an error or not. If it doesn't pass then again you certainly need to inspect the current oracle to check what's the issue with the output.
For test_clipping though I think Brennan encoded these error cases with a
e_
prefix (instead of this.error
suffix)Anyway, I'd close this ticket as _wontfix_, ok with you Bang?