On 10/14/11 6:13 AM, Kristian Waagan wrote:
On 14.10.11 14:28, Rick Hillegas wrote:
On 10/14/11 1:38 AM, Kristian Waagan wrote:
On 10/13/11 02:47 PM, Rick Hillegas wrote:
On 10/13/11 3:12 AM, Kristian Waagan wrote:
On 10/13/11 12:48 AM, Kathey Marsden wrote:
FYI, I filed this IFRA issue which has some impact on the Derby
Jira VTI in our demos.
I suppose the demo will still work but now Jira export is limited
to 1000 issues. Is a DERBY issue in order as well?
Working around that limit is possible in this case [1] - the
question is if we want to do that?
If people are using this demo extensively as their own private
"Derby JIRA cache", I think some effort should be invested into
writing something that imposes less load on the ASF JIRA instance.
Writing something solid here may require quite a bit of effort.
Without having given this extensive thought, I think parsing emails
as suggested on infra is the kindest option. This would probably
suit committers well.
Maybe the activity stream (as an RSS feed) could be used too, but in
that case it may be harder to make sure no updates are missed.
A third option is to use the SOAP API. One has to be able to define
a suitable search (for instance 'give me all updates on Derby from
date X to date Y'), but I suspect you'd have to download all the
issue information instead of just what changed in that period. The
upside with this approach compared to the RSS one, is that you
decide when you want to update your copy.
Now that we (by which I mean Kristian) know how to use the SOAP API,
it may be possible to write a restricted vti that wraps the SOAP API.
I believe that the existing vti is used to dramatically reduce the
set of issues you have to look at by applying restrictions which
cannot be expressed in the JIRA gui.
What are examples of such restrictions?
Hi Kristian,
Here is a sample query which uses the existing jira xml vti (from
http://wiki.apache.org/db-derby/HowToRunJiraVtiReports):
select s."key", s."assignee", s."updated" , s."title"
from table( "apacheVanillaJiraReport"(
'file:///kmarsden/projects/jirareports/all_JIRA_ISSUES.xml' ) ) s where
"resolution" = 'Unresolved' and "assignee" != 'Unassigned' and NOT
"updated" like '%2008%' order by "assignee";
Thanks, Rick.
This translates roughly to the following JQL (I changed the date):
project = 'DERBY' AND resolution = 'Unresolved' AND assignee is not
NULL AND updated < '2011-01-01'
(returns 25 issues by the way)
There are two interesting things about this query:
1) Projection (fancy term for "SELECT list")- The SELECT list is just a
few of the columns in the vti. That is, the width of the query result is
considerably narrower than the full width of the vti.
2) Restriction (fancy term for pieces of the WHERE clause) - A couple
terms in the WHERE clause could be pushed into the Store if the query
were directed against a Derby table rather than a vti. These terms
reduce the length of the result set coming back to the SQL interpreter.
These terms are:
"resolution" = 'Unresolved'
"assignee" != 'Unassigned'
If the query were issued against a restricted vti which wrapped the SOAP
api, then the projection and the restriction could be pushed into the
vti. That is, Derby would tell the vti to fetch only 4 columns and to
throw away any rows which didn't match the two WHERE clause terms above.
Presumably, this information could be passed through the SOAP api to
limit the width and length of the result set being requested from JIRA.
The projection cannot be used to download less data with the SOAP API.
The restrictions would be helpful though, but then we need a SQL ->
JQL translator, or we can cheat and accept the JQL as an argument to
the VTI.
I haven't looked at how the VTI restrictions are represented, but for
the sake of a demo it should be feasible to write some code that
converts the VTI restrictions into JQL.
Hi Kristian,
The restriction is represented as a org.apache.derby.vti.Restriction.
That class has a toSQL() method which turns the restriction into SQL
suitable for sending to a foreign rdbms. JQL may be enough like SQL that
the existing toSQL() output will be good enough. If not, turning a
Restriction into a JQL WHERE clause should be straightforward (tm).
But if you can express your query using JQL, why not go directly to JIRA?
I think that the problem is that there are some queries (particularly
those involving NOT) which can't be constructed using the existing JIRA
gui. You can say "I want all issues with the following fields turned on"
but you can't say "I want all issues with the following field NOT turned
on".
Thanks,
-Rick
I see the point of caching and dataset rip-offs (i.e. you populate a
table and continue to refine the results/searches locally), are there
other use-cases where people think they would prefer Derby+VTI over
JIRA+JQL?
