Skip to content
New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Improved execution field merging #2019

Merged
merged 4 commits into from
Nov 25, 2023
Merged

Improved execution field merging #2019

Changes from 3 commits
Commits
Show all changes
4 commits
Select commit Hold shift + click to select a range
fa95ccb
Avoid unnecessary field merging during execution
kyri-petrou Nov 23, 2023
3e7ec5a
Cleanup
kyri-petrou Nov 23, 2023
547efa2
PR comments
kyri-petrou Nov 23, 2023
c6201fe
Extract Map size calculation into separate method
kyri-petrou Nov 24, 2023
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
52 changes: 34 additions & 18 deletions core/src/main/scala/caliban/execution/Executor.scala
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -306,27 +306,43 @@ object Executor {
ZIO.succeed(GraphQLResponse(NullValue, List(error)))

private[caliban] def mergeFields(field: Field, typeName: String): List[Field] = {
val map = new java.util.LinkedHashMap[String, Field]()
var modified = false

field.fields.foreach { field =>
if (field._condition.forall(_.contains(typeName))) {
map.compute(
field.aliasedName,
(_, f) =>
if (f == null) field
else {
modified = true
f.copy(fields = f.fields ::: field.fields)
}
)
} else {
modified = true
def haveSameCondition(head: Field, tail: List[Field]): Boolean = {
val condition = head._condition
var remaining = tail
while (!remaining.isEmpty) {
if (remaining.head._condition != condition) return false
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I was gonna suggest tail recursion but this will probably iterate tighter

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Hmm I think the compiler will rewrite the tailrec method likely to this implementation. I'll check, if it's the same I'll use tailrec

Copy link
Collaborator Author

@kyri-petrou kyri-petrou Nov 24, 2023 •
edited
Loading

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Looking at the generated code, it looks like @tailrec uses Java labels in an infinite while loop with return statements, whereas this one uses the !remaining.isEmpty method as a condition to terminate the while loop. Not sure which one would be more performant in theory

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Yeah we're probably getting into a grey area of compiler optimization. I think the conditionless while might be able to skip a branch if equal instruction, though for modern chips branches aren't really performance killers anymore

remaining = remaining.tail
}
true
}

// Avoid conversions if no modification took place
if (modified) map.values().asScala.toList else field.fields
def matchesTypename(f: Field): Boolean =
f._condition.isEmpty || f._condition.get.contains(typeName)

def mergeFields(fields: List[Field]) = {
val map = new java.util.LinkedHashMap[String, Field]((fields.size / 0.75d).toInt + 1)
Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Is this a known formula? Or just an estimate?

Copy link
Owner

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Would be good to explain what this is in a comment

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The exact formula would be Math.ceil(size / 0.75) but this is close enough I believe. I can use that instead if you prefer.

The Map resizes when the number of items in it exceed the capacity * loading factor (defaults to 0.75). Using this formula guarantees that we have adequate capacity so that the Map won't need to resize. In JDK17+ there is a method that allows you to create the Maps taking the expected number of items as a parameter (instead of using the initial size) and it used this formula.

I'll add a comment for it.

Copy link
Collaborator

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

I forget is scala list's size O(n) or is it stored on concat?

Copy link
Collaborator Author

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

It's O(n), although given that this List only contains the queried fields and that this method is only run for objects with multiple fragments, I think the overall impact of counting the number of elements in this list is extremely small

var remaining = fields
while (!remaining.isEmpty) {
val h = remaining.head
if (matchesTypename(h)) {
map.compute(
h.aliasedName,
(_, f) =>
if (f eq null) h
else f.copy(fields = f.fields ::: h.fields)
)
}
remaining = remaining.tail
}
map.values().asScala.toList
}

field.fields match {
// Shortcut if all the fields have the same condition, which means we don't need to merge as that's been handled in Field.apply
case h :: t if haveSameCondition(h, t) => if (matchesTypename(h)) field.fields else Nil
case Nil => Nil
case fields => mergeFields(fields)
}
}

private def fieldInfo(field: Field, path: List[Either[String, Int]], fieldDirectives: List[Directive]): FieldInfo =
Expand Down