Not "reorder these 5 rules". Rules are matched by patterns, not orders. What I mean is to only use one heuristic rule, and remove the default commute/associate rule. Programs#heuristicJoinOrder[1] is an example to do so.
[1] https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/tools/Programs.java#L210-L219 Suchit Kumar <[email protected]> 于2023年2月9日周四 11:07写道： > Hi Benchao, > > So If I am able to understand correctly you are saying if we are able to > reorder these 5 rules then this issue may resolve right? > > Thanks, > Suchit Kumar > > On Thu, Feb 9, 2023 at 8:27 AM Benchao Li <[email protected]> wrote: > > > Besides what Stamatis and Mihai mentioned, I'd love to share my current > > insight for your case. > > > > In this example, you have 6 tables joining together, the problem seems > > related with the join reordering optimizations. > > > > The default join reorder rules are JOIN_COMMUTE[1] and JOIN_ASSOCIATE[2]. > > There are other heuristic join reorder rules such as > > MULTI_JOIN_OPTIMIZE_BUSHY[3], MULTI_JOIN_OPTIMIZE[4], which are more > > efficient in complex join scenarios. If you could managed to custom your > > optimizing rules, you can try these rules. > > > > There is a utility method to enable this[5] which is only used in tests, > it > > will be helpful if we also have a configuration for this behavior. > > > > [1] > > > > > https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/rel/rules/JoinCommuteRule.java#L57 > > [2] > > > > > https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/rel/rules/JoinAssociateRule.java#L51 > > [3] > > > > > https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/rel/rules/MultiJoinOptimizeBushyRule.java#L78 > > [4] > > > > > https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/rel/rules/LoptOptimizeJoinRule.java#L77 > > [5] > > > > > https://github.com/apache/calcite/blob/b64cb1325cfe1a5143ea3ca534f991b6f881c3c5/core/src/main/java/org/apache/calcite/tools/Programs.java#L182-L189 > > > > Mihai Budiu <[email protected]> 于2023年2月9日周四 06:06写道： > > > > > You haven't described your setup precisely enough for me to understand > > > what is going on. > > > But in general with Calcite you can control very precisely which plan > > > rewrite rules are being applied and when. > > > > > > Here is what we do. > > > In the SqlToRelConverter we use a HepPlanner with an empty set of > rules. > > > After we get a RelNode from the converter we run a series of > HepPlanners, > > > each with a list of hand-picked rules. So we control exactly what code > > > transformations are applied. > > > In fact, we can even look at the query code and choose the list > > > transformations depending on that; for example, the BushyJoin > > optimization > > > rules does not work for outer joins, so if the plan contains outer > joins, > > > we avoid applying this rule. > > > Here is the source code for our current workflow: > > > > > > https://github.com/vmware/sql-to-dbsp-compiler/blob/main/SQL-compiler/src/main/java/org/dbsp/sqlCompiler/compiler/sqlparser/CalciteCompiler.java > > > > > > I am relatively new to Calcite, so I may not be using it in the best > way, > > > but I think it gives you complete control. > > > For infinite loops, perhaps you could setup a timer and abort > executing a > > > optimization step when it expires. > > > Mihai > > > > > > > > > > -----Original Message----- > > > > From: Stamatis Zampetakis <[email protected]> > > > > Sent: Wednesday, February 8, 2023 1:52 PM > > > > To: [email protected] > > > > Subject: Re: Volcano Planner Issue > > > > > > > > !! External Email > > > > > > > > You cannot execute a query without planning it. I don't have enough > > > context > > > > to answer your question. Sorry! > > > > > > > > Best, > > > > Stamatis > > > > > > > > On Wed, Feb 8, 2023, 7:22 PM Suchit Kumar <[email protected]> > > > > wrote: > > > > > > > > > Hi Stamatis, > > > > > > > > > > Hope you are doing well. > > > > > I will upload this to git meanwhile can you suggest if is there any > > > > > way we can bypass the planner and directly execute the query? > > > > > > > > > > Thanks, > > > > > Suchit Kumar > > > > > > > > > > On Wed, Feb 8, 2023 at 3:57 PM Stamatis Zampetakis > > > > <[email protected]> > > > > > wrote: > > > > > > > > > > > Hey Suchit, > > > > > > > > > > > > It is usually better to share links to GiHub (or other CVS > > > > > > platforms) > > > > > with > > > > > > projects showing the problem instead of attaching zip files. > > > > > > Many security conscious people will never download a zip file > from > > > > > someone > > > > > > they don't know. > > > > > > Moreover, with GitHub it is easier for people to give a quick > > glance > > > > > > in case they can spot something quickly. > > > > > > > > > > > > If you need help with Calcite, this list is the best place to get > > > help. > > > > > > Take into account that we are all volunteers with limited time so > > > > > > there > > > > > is > > > > > > no guarantee that someone will pick this up. > > > > > > > > > > > > Best, > > > > > > Stamatis > > > > > > > > > > > > On Wed, Feb 8, 2023 at 3:31 AM Suchit Kumar < > [email protected] > > > > > > > > wrote: > > > > > > > > > > > > > Hi, > > > > > > > > > > > > > > > > > > > > > Thanks for your reply could you guide me if there is any way I > > can > > > > > > execute > > > > > > > my query without an optimizer? > > > > > > > I am not sure whom I can reach can you tell me any channel > where > > I > > > > > > > can > > > > > > post > > > > > > > this issue? > > > > > > > > > > > > > > Thanks, > > > > > > > Suchit Kumar > > > > > > > > > > > > > > > > > > > > > On Wed, Feb 8, 2023 at 7:57 AM Benchao Li < > [email protected]> > > > > > wrote: > > > > > > > > > > > > > > > Hi Suchit, > > > > > > > > > > > > > > > > FYI, I've reproduced the problem with your example, however, > I > > > > > > > > cannot > > > > > > > tell > > > > > > > > the reason for now, it may take a while. > > > > > > > > > > > > > > > > I would be great if someone else who's interested with the > > > > > > > > problem > > > > > > could > > > > > > > > also take a look at it. > > > > > > > > > > > > > > > > Suchit Kumar <[email protected]> 于2023年2月7日周二 11:17 > > > > 写道： > > > > > > > > > > > > > > > > > Hi, > > > > > > > > > > > > > > > > > > I have attached 4 files you can run ApacheClient file to > > check > > > > > > > > > the example. Please let me know if you face any issues > > running > > > > > > > > > my > > > > > > example. > > > > > > > > > > > > > > > > > > Thanks, > > > > > > > > > Suchit Kumar > > > > > > > > > > > > > > > > > > On Tue, Feb 7, 2023 at 7:31 AM Benchao Li > > > > > > > > > <[email protected]> > > > > > > > wrote: > > > > > > > > > > > > > > > > > >> Hi Suchit, > > > > > > > > >> > > > > > > > > >> Sorry that I couldn't find your example. Am I missing > > > something? > > > > > > > > >> > > > > > > > > >> Suchit Kumar <[email protected]> 于2023年2月7日周二 > > > > 00:01写道： > > > > > > > > >> > > > > > > > > >> > Hi, > > > > > > > > >> > > > > > > > > > >> > Thank you for your reply, This is the example that I am > > > > > > > > >> > creating > > > > > > and > > > > > > > > >> > inside iterativeRuleQueue it keeps running. > > > > > > > > >> > In this example, I need to query on top of the java > > hashmap. > > > > > > > > >> > Please let me know what I am missing and if there is > > > > > > > > >> > anyway. So > > > > > > > that I > > > > > > > > >> can > > > > > > > > >> > bypass the optimiser and run the query directly. > > > > > > > > >> > > > > > > > > > >> > > > > > > > > > >> > > > > > > > > > >> > On Mon, Feb 6, 2023 at 4:39 PM Benchao Li > > > > > > > > >> > <[email protected] > > > > > > > > > > > > > > wrote: > > > > > > > > >> > > > > > > > > > >> >> Hi Suchit, > > > > > > > > >> >> > > > > > > > > >> >> It would be helpful if you could provide more > information > > > > > > > > >> >> such > > > > > > as: > > > > > > > > >> >> - What the version are you using > > > > > > > > >> >> - How you used Calcite, e.g., what's the rules you are > > > > > > > > >> >> using, > > > > > is > > > > > > > > there > > > > > > > > >> any > > > > > > > > >> >> custom rule? > > > > > > > > >> >> - If there is a pure Calcite reproducible test case, it > > > > > > > > >> >> would > > > > > be > > > > > > > > easier > > > > > > > > >> >> for > > > > > > > > >> >> others to understand the problem and propose a solution > > > > > > > > >> >> > > > > > > > > >> >> Come back to your question, I'm guessing that is there > a > > > > > > > > >> >> custom > > > > > > > rule > > > > > > > > >> which > > > > > > > > >> >> is not implemented well, and it always matches and > > > > > > > > >> >> produces > > > > > > another > > > > > > > > >> result > > > > > > > > >> >> which could match the rule again, which will go into an > > > > > infinite > > > > > > > > loop? > > > > > > > > >> >> > > > > > > > > >> >> Suchit Kumar <[email protected]> 于2023年2月6日周一 > > > > 13:13写道： > > > > > > > > >> >> > > > > > > > > >> >> > Hi Team, > > > > > > > > >> >> > > > > > > > > > >> >> > I am using apache calcite for one of our use cases > but > > > > > > > > >> >> > in 1 > > > > > > > query I > > > > > > > > >> am > > > > > > > > >> >> > getting an issue where the volcano planner is not > > > > > > > > >> >> > returning > > > > > > > > anything > > > > > > > > >> >> and it > > > > > > > > >> >> > keeps running I have checked online as well but I > > > > > > > > >> >> > haven't > > > > > found > > > > > > > any > > > > > > > > >> >> > solution. Could you please let me know if there is > any > > > > > > > > >> >> > way to > > > > > > > > disable > > > > > > > > >> >> the > > > > > > > > >> >> > planner optimizer? > > > > > > > > >> >> > > > > > > > > > >> >> > Thanks, > > > > > > > > >> >> > Suchit Kumar > > > > > > > > >> >> > > > > > > > > > >> >> > > > > > > > > >> >> > > > > > > > > >> >> -- > > > > > > > > >> >> > > > > > > > > >> >> Best, > > > > > > > > >> >> Benchao Li > > > > > > > > >> >> > > > > > > > > >> > > > > > > > > > >> > > > > > > > > >> -- > > > > > > > > >> > > > > > > > > >> Best, > > > > > > > > >> Benchao Li > > > > > > > > >> > > > > > > > > > > > > > > > > > > > > > > > > > -- > > > > > > > > > > > > > > > > Best, > > > > > > > > Benchao Li > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > > !! External Email: This email originated from outside of the > > > organization. Do > > > > not click links or open attachments unless you recognize the sender. > > > > > > > > > -- > > > > Best, > > Benchao Li > > > -- Best, Benchao Li
