Enhanced signature generation #5153
Conversation
This reverts commit 3c71ca9.
|
CI performance tests
|
BrynCooke
force-pushed
the
njm/P-695/sig-gen-algo
branch
from
b76d8b5
to
d17c66b
Compare
| @@ -356,33 +404,60 @@ fn format_selection_set(selection_set: &SelectionSet, f: &mut fmt::Formatter) -> | |||
| } | |||
|
|
|||
| if !fields.is_empty() || !named_fragments.is_empty() || !inline_fragments.is_empty() { | |||
| fields.sort_by(|&a, &b| a.name.cmp(&b.name)); | |||
| if is_enhanced(normalization_algorithm) { | |||
There was a problem hiding this comment.
Why don't we just call it once for everywhere in the code instead of recalling this function everytime ?
There was a problem hiding this comment.
We'd have to pass around a boolean instead of the enum if we only wanted to call this once, and I figured it would be better to pass around the enum instead in case we add a new algorithm in the future. I'll change it to a macro though which should fix any performance concerns.
| f.write_str("{")?; | ||
| for (index, (name, val)) in o.iter().enumerate() { | ||
| if index != 0 { | ||
| f.write_str(",")?; | ||
| } | ||
| write!(f, "{}:", name)?; | ||
| format_value(val, normalization_algorithm, f)?; | ||
| } | ||
| f.write_str("}") |
There was a problem hiding this comment.
Are we basically trying to serialize it in json here ?
There was a problem hiding this comment.
It's similar but with enough special cases to justify not using a JSON serializer (IMO). We need to strip the string and int values and the array contents, and also the keys are not surrounded in quotes.
| // We need to use this because of the import method that the fuzzer uses. The fuzzer can't have apollo-router as | ||
| // a direct dependency because that means the build crashes, so we use a dev dependency and do a path hack. This | ||
| // means that the fuzzer can't import from the Configuration namespace and has to import this instead. | ||
| #[allow(dead_code)] |
There was a problem hiding this comment.
When I remove it I get the error variant ``Enhanced`` is never constructed even though it's obviously used below. I'm not sure if there's a better way to get around this rather than having the #[allow(dead_code)]?
| #[derive(Clone, PartialEq, Eq, Default, Derivative, Serialize, Deserialize, JsonSchema)] | ||
| #[derivative(Debug)] | ||
| #[serde(rename_all = "lowercase")] | ||
| pub(crate) enum ApolloSignatureNormalizationAlgorithm { |
There was a problem hiding this comment.
Why do we have a duplicate enum here and in the studio_interop module ?
There was a problem hiding this comment.
This was a consequence of the way we had to do the weird import of the module in the fuzz test:
#[path = "../../apollo-router/src/apollo_studio_interop/mod.rs"]
mod apollo_router_usage_reporting;
use apollo_router_usage_reporting::generate_usage_reporting;
use apollo_router_usage_reporting::SignatureNormalizationAlgorithm;
use apollo_router_usage_reporting::UsageReportingComparisonResult;
Because of that I was unable to use the enum defined in the configuration module in the interop module. I had a DM conversation with Jeremy and Simon and the separate enums was the best workaround we could think of.
There's another thread on why we needed to do the import that way somewhere and I try to can find it if you're interested but the end result was that we just had to do it that way.
| if self.experimental_apollo_signature_normalization_algorithm | ||
| == ApolloSignatureNormalizationAlgorithm::Enhanced | ||
| && self.experimental_apollo_metrics_generation_mode != ApolloMetricsGenerationMode::New | ||
| { | ||
| return Err(ConfigurationError::InvalidConfiguration { | ||
| message: "`experimental_apollo_signature_normalization_algorithm: enhanced` requires `experimental_apollo_metrics_generation_mode: new`", | ||
| error: "either change to the legacy signature normalization mode, or change to new metrics generation".into() | ||
| }); | ||
| } |
There was a problem hiding this comment.
We should add a test covering this rule because once we will update the configuration field without the experimental support we will also want to update this error message
| let signature_normalization_mode = match self | ||
| .configuration | ||
| .experimental_apollo_signature_normalization_algorithm | ||
| { | ||
| ApolloSignatureNormalizationAlgorithm::Legacy => { | ||
| SignatureNormalizationAlgorithm::Legacy | ||
| } | ||
| ApolloSignatureNormalizationAlgorithm::Enhanced => { | ||
| SignatureNormalizationAlgorithm::Enhanced | ||
| } | ||
| }; |
There was a problem hiding this comment.
If we still have to duplicate this enum maybe a From impl would be better
|
I've addressed the PR comments but I don't have access to push to this branch/PR. I made a new branch/PR but it's based on the latest dev code so there are some unrelated changes in the PR. If this branch is synced with dev I think the PR should look a lot cleaner. |
Adds a new experimental configuration to support enhanced signature generation. The changes are:
Remove some hacks and quirks of the JS implementation related to properly sorting everything and using commas vs spaces
Includes aliases in the signature
Includes the full normalized form of input objects
Fixes ROUTER-262
Replaces PR #5062
To get this over the line the tests need a little refactoring to use insta and maybe Asset to allow easy addition of new tests.
Checklist
Complete the checklist (and note appropriate exceptions) before the PR is marked ready-for-review.
Exceptions
Note any exceptions here
Notes
Footnotes
It may be appropriate to bring upcoming changes to the attention of other (impacted) groups. Please endeavour to do this before seeking PR approval. The mechanism for doing this will vary considerably, so use your judgement as to how and when to do this. ↩
Configuration is an important part of many changes. Where applicable please try to document configuration examples. ↩
Tick whichever testing boxes are applicable. If you are adding Manual Tests, please document the manual testing (extensively) in the Exceptions. ↩