Post validation transformations

[aew]

Various bugfixes and optimizations:

• Add test for execution on field arguments with variable bindings
• Build arguments, expand fragments, and inline types prior to execution phase
• Refactor executor to be simpler (delete the majority of the code)
• Warnings for library users accidentally doing parsing / validation in the inner loop
• Bugfix - break on undefined variables, don't traverse deeper
• Bugfix - add test for missing required variables, and support default values for variables
• Bugfix - guard against null values for non-null types in the execution phase
• Bugfix - support execution with inline fragments
• Bugfix - validation for missing root fields
• Bugfix - required arguments fulfilled by required variables
• Add preparation and memoization as an example in executor test
• Optionally add a :resolver key to state when validating a statement to inline resolver-fns prior to execution (and add example to executor test case)
• Streamline statement validation output - remove :state key in the absence of errors
• Add a spec for statement validation output and ensure test cases conform
• Add starwars test case from the graphql-clj-starter project
• Bugfix for introspection resolution - isDeprecated is a non-null field

Addresses:

• https://github.com/tendant/graphql-clj/issues/42
• https://github.com/tendant/graphql-clj/issues/36 - doesn't change the current approach, but removes much of the complexity from it
• https://github.com/tendant/graphql-clj/issues/35
• https://github.com/tendant/graphql-clj/issues/43 - partial fix - relaxes validation constraint for now until we can resolve object equality for these specs - better than blocking valid statements in the meantime.

Related:

• https://github.com/tendant/graphql-clj-starter/pull/1
• https://github.com/xsc/graphql-clj-bench/pull/2

[aew]

Validated statement output should now be roughly equivalent in size to the parsed data structure

[aew]

The changes to the executor in this PR have the following impact on execution runtime performance:

version 1.19.0 with memoization

Evaluation count : 1479300 in 60 samples of 24655 calls.
             Execution time mean : 41.074870 µs
    Execution time std-deviation : 649.450286 ns
   Execution time lower quantile : 40.055590 µs ( 2.5%)
   Execution time upper quantile : 42.387063 µs (97.5%)

version 1.20.0-SNAPSHOT with memoization

Evaluation count : 4248660 in 60 samples of 70811 calls.
             Execution time mean : 14.944263 µs
    Execution time std-deviation : 575.873227 ns
   Execution time lower quantile : 14.091584 µs ( 2.5%)
   Execution time upper quantile : 16.059733 µs (97.5%)

version 1.20.0-SNAPSHOT with memoization and inline resolver functions

Evaluation count : 6908580 in 60 samples of 115143 calls.
             Execution time mean : 8.915279 µs
    Execution time std-deviation : 160.062685 ns
   Execution time lower quantile : 8.687222 µs ( 2.5%)
   Execution time upper quantile : 9.306784 µs (97.5%)

From: https://github.com/xsc/graphql-clj-bench/pull/2

[tendant]

Found a bug: https://github.com/tendant/graphql-clj-starter/tree/testing-validation-loc

Required List field has wrong kind :OBJCET instead of :LIST.

[xsc]

@tendant Stuff like that should probably be verified within the inteospection unit tests, right?

[aew]

I've fixed this bug, and added some unit test cases.

[tendant]

Super. I will test it and merge it soon.