Hi Dave,

The "I" (Capital i) button in display should work. ItemExpr tree can also
be seen by right clicking on a RelExpr and choosing ItemExpr from the drop
down menu (three choices in the menu I think). One can oonly see ItemExpr
trees at the end of major compiler phases though with DISPLAY. Also maybe
not all ItemExpr trees of a RelExpr are displayed, only the ones the tool
is is aware of.

Thanks
Suresh

On Tue, Nov 3, 2015 at 6:53 PM, Hans Zeller <[email protected]> wrote:

> Hi Dave,
>
> Many years ago, we used to have a method ExprNode::displayTree() that would
> invoke the GUI display tool from the debugger, so you could look at a tree
> not just at the end of each phase but also in-between. It looks like the
> current GUI does not support this feature, but maybe we could try to bring
> it back. I don't know how much work that would be.
>
> Hans
>
> On Tue, Nov 3, 2015 at 3:47 PM, Dave Birdsall <[email protected]>
> wrote:
>
> > Hi,
> >
> >
> >
> > I’ve been debugging some issues in the Trafodion compiler where the fix
> > involves transforming the query tree in a certain way. One issue I’m
> > running into is getting the transformation right. What would be really
> > helpful is some utility that would show the details of a tree (the
> RelExpr
> > nodes and all the dangling ItemExpr subtrees under it, complete with
> > ValueIDs) so I could compare my transformation with other
> transformations;
> > that is, so I can get a visual picture of what a good transformation
> looks
> > like.
> >
> >
> >
> > I have been using the DISPLAY utility, but it doesn’t seem to show the
> > ItemExpr subtrees. Perhaps there’s a bug in there that needs to be fixed?
> >
> >
> >
> > The ItemExpr nodes have a display() method that is handy to call from
> gdb.
> > If I could have something like that on RelExpr nodes as well, and with
> > ValueIDs added in all the ItemExpr nodes, that would be great, I think.
> >
> >
> >
> > Perhaps this capability exists and I just haven’t stumbled upon it yet?
> >
> >
> >
> > Thanks for any suggestions,
> >
> >
> >
> > Dave
> >
>

Reply via email to