mirror of
https://github.com/github/codeql.git
synced 2025-12-18 01:33:15 +01:00
a73e382dfe
I'm not entirely sure what triggered this bad join order, but some
combination of the use of abstract classes and the exclusion of `new`
caused this to go really wrong:
```
WeakSensitiveDataHashing.ql-15:Stdlib::Stdlib::HashlibDataPassedToHashClass#class#ffff ......... 15.5s
```
with the following tuple counts:
```
[2021-07-12 13:20:15] (16s) Tuple counts for Stdlib::Stdlib::HashlibDataPassedToHashClass#class#ffff/4@217901:
148810 ~3% {3} r1 = JOIN DataFlowPublic::CallCfgNode#class#ff#shared WITH project#DataFlowPublic::CallCfgNode::getArg_dispred#fff ON FIRST 1 OUTPUT "hashlib", Lhs.1 'node', Lhs.0 'this'
148810 ~4% {3} r2 = JOIN r1 WITH ApiGraphs::API::Impl::MkModuleImport#ff@staged_ext ON FIRST 1 OUTPUT Rhs.1, Lhs.1 'node', Lhs.2 'this'
7589310 ~486% {4} r3 = JOIN r2 WITH ApiGraphs::API::Impl::edge#2#fff@staged_ext ON FIRST 1 OUTPUT Lhs.1 'node', Lhs.2 'this', Rhs.1, InverseAppend("getMember(\"","\")",Rhs.1)
6994070 ~490% {4} r4 = SELECT r3 ON In.3 != "new"
6994070 ~4503% {2} r5 = SCAN r4 OUTPUT In.1 'this', In.0 'node'
22 ~4% {3} r6 = JOIN DataFlowPublic::CallCfgNode#class#ff#shared WITH project#DataFlowPublic::CallCfgNode::getArgByName_dispred#fff ON FIRST 1 OUTPUT "hashlib", Lhs.1 'node', Lhs.0 'this'
22 ~0% {3} r7 = JOIN r6 WITH ApiGraphs::API::Impl::MkModuleImport#ff@staged_ext ON FIRST 1 OUTPUT Rhs.1, Lhs.1 'node', Lhs.2 'this'
1122 ~437% {4} r8 = JOIN r7 WITH ApiGraphs::API::Impl::edge#2#fff@staged_ext ON FIRST 1 OUTPUT Lhs.1 'node', Lhs.2 'this', Rhs.1, InverseAppend("getMember(\"","\")",Rhs.1)
1034 ~460% {4} r9 = SELECT r8 ON In.3 != "new"
1034 ~4549% {2} r10 = SCAN r9 OUTPUT In.1 'this', In.0 'node'
6995104 ~4503% {2} r11 = r5 UNION r10
5213851 ~4683% {3} r12 = JOIN r11 WITH ApiGraphs::API::Node::getACall_dispred#ff_10#join_rhs ON FIRST 1 OUTPUT Rhs.1 'hashClass', Lhs.1 'node', Lhs.0 'this'
6478480 ~4646% {6} r13 = JOIN r12 WITH ApiGraphs::API::Impl::edge#2#fff_201#join_rhs ON FIRST 1 OUTPUT "hashlib", Rhs.1, Lhs.1 'node', Lhs.2 'this', Lhs.0 'hashClass', Rhs.2
1410 ~4693% {5} r14 = JOIN r13 WITH ApiGraphs::API::Impl::MkModuleImport#ff@staged_ext ON FIRST 2 OUTPUT Lhs.2 'node', Lhs.3 'this', Lhs.4 'hashClass', Lhs.5, InverseAppend("getMember(\"","\")",Lhs.5)
1222 ~4540% {5} r15 = SELECT r14 ON In.4 'hashName' != "new"
1222 ~4540% {4} r16 = SCAN r15 OUTPUT In.1 'this', In.4 'hashName', In.2 'hashClass', In.0 'node'
```
By factoring out the insides, the biggest iteration now looks like
```
[2021-07-12 14:17:36] (0s) Tuple counts for Stdlib::Stdlib::HashlibDataPassedToHashClass#class#ffff/4@85bb21:
148810 ~0% {2} r1 = JOIN DataFlowPublic::CallCfgNode#class#ff#shared WITH project#DataFlowPublic::CallCfgNode::getArg_dispred#fff ON FIRST 1 OUTPUT Lhs.1 'node', Lhs.0 'this'
148810 ~0% {2} r2 = JOIN r1 WITH Stdlib::Stdlib::hashlibMember#ff#nonempty CARTESIAN PRODUCT OUTPUT Lhs.1 'this', Lhs.0 'node'
22 ~0% {2} r3 = JOIN DataFlowPublic::CallCfgNode#class#ff#shared WITH project#DataFlowPublic::CallCfgNode::getArgByName_dispred#fff ON FIRST 1 OUTPUT Lhs.1 'node', Lhs.0 'this'
22 ~0% {2} r4 = JOIN r3 WITH Stdlib::Stdlib::hashlibMember#ff#nonempty CARTESIAN PRODUCT OUTPUT Lhs.1 'this', Lhs.0 'node'
148832 ~0% {2} r5 = r2 UNION r4
110933 ~2% {3} r6 = JOIN r5 WITH ApiGraphs::API::Node::getACall_dispred#ff_10#join_rhs ON FIRST 1 OUTPUT Rhs.1 'hashClass', Lhs.1 'node', Lhs.0 'this'
26 ~0% {4} r7 = JOIN r6 WITH Stdlib::Stdlib::hashlibMember#ff_10#join_rhs ON FIRST 1 OUTPUT Lhs.2 'this', Rhs.1 'hashName', Lhs.0 'hashClass', Lhs.1 'node'
return r7
```
(The tuple counts themselves are not directly comparable.)