[ 
https://issues.apache.org/jira/browse/CALCITE-4737?page=com.atlassian.jira.plugin.system.issuetabpanels:comment-tabpanel&focusedCommentId=17459849#comment-17459849
 ] 

Thomas Rebele edited comment on CALCITE-4737 at 12/15/21, 10:50 AM:
--------------------------------------------------------------------

[~zabetak] had proposed to make the visualizer independent of the 
VolcanoPlanner. As this would entail many chances to the Java part, I decided 
to solve at the same time another problem that I had encountered in practice: 
the size of the generated files is about Θ(nr), with n the number of nodes and 
r the number of rule calls. Instead of storing the state at each rule 
application, the new format just stores the changes at each step. So the size 
of new format should be closer to Θ(n + r). The visualizer now works with the 
VolcanoPlanner and the HepPlanner.

[~Chunwei Lei], [~zuozhiw], and [~zabetak] proposed to add a property to 
activate generating the files for the visualizer, maybe depending on the log 
level. How about adding a property to switch between printing the dot graph (as 
it is currently done) and the visualizer? I would add a property for specifying 
the output directory and the file name prefix, include a counter to the 
filename (and maybe a timestamp), and print the name of the generated file in 
the log. Any other proposals?


was (Author: thomas.rebele):
[~zabetak] had proposed to make the visualizer independent of the 
VolcanoPlanner. As this would entail many chances to the Java part, I decided 
to solve at the same time another problem that I had encountered in practice: 
the size of the generated files is about Θ(nr), with n the number of nodes and 
r the number of rule calls. Instead of storing the state at each rule 
application, the new format just stores the changes at each step. So the size 
of new format should be closer to Θ(n + r). The visualizer now works with the 
VolcanoPlanner and the HepPlanner.

[~Chunwei Lei] and [~zabetak] proposed to add a property to activate generating 
the files for the visualizer, maybe depending on the log level. How about 
adding a property to switch between printing the dot graph (as it is currently 
done) and the visualizer? I would add a property for specifying the output 
directory and the file name prefix, include a counter to the filename (and 
maybe a timestamp), and print the name of the generated file in the log. Any 
other proposals?

> Add Volcano visualizer for debugging
> ------------------------------------
>
>                 Key: CALCITE-4737
>                 URL: https://issues.apache.org/jira/browse/CALCITE-4737
>             Project: Calcite
>          Issue Type: Bug
>            Reporter: Julian Hyde
>            Priority: Major
>              Labels: pull-request-available
>          Time Spent: 7h 10m
>  Remaining Estimate: 0h
>
> Add Volcano visualizer for debugging.



--
This message was sent by Atlassian Jira
(v8.20.1#820001)

Reply via email to