The simplification will allow us to avoid a few `MonadError QErr m` constsraints in the future - this effect can be created locally instead of reusing a global one.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/8729
GitOrigin-RevId: 851e28b1f5bfe4c47da43fa324714a941ef25c57
This PR is on top of #7789.
### Description
This PR entirely rewrites the API of the Tracing library, to make `interpTraceT` a thing of the past. Before this change, we ran traces by sticking a `TraceT` on top of whatever we were doing. This had several major drawbacks:
- we were carrying a bunch of `TraceT` across the codebase, and the entire codebase had to know about it
- we needed to carry a second class constraint around (`HasReporterM`) to be able to run all of those traces
- we kept having to do stack rewriting with `interpTraceT`, which went from inconvenient to horrible
- we had to declare several behavioral instances on `TraceT m`
This PR rewrite all of `Tracing` using a more conventional model: there is ONE `TraceT` at the bottom of the stack, and there is an associated class constraint `MonadTrace`: any part of the code that happens to satisfy `MonadTrace` is able to create new traces. We NEVER have to do stack rewriting, `interpTraceT` is gone, and `TraceT` and `Reporter` become implementation details that 99% of the code is blissfully unaware of: code that needs to do tracing only needs to declare that the monad in which it operates implements `MonadTrace`.
In doing so, this PR revealed **several bugs in the codebase**: places where we were expecting to trace something, but due to the default instance of `HasReporterM IO` we would actually not do anything. This PR also splits the code of `Tracing` in more byte-sized modules, with the goal of potentially moving to `server/lib` down the line.
### Remaining work
This PR is a draft; what's left to do is:
- [x] make Pro compile; i haven't updated `HasuraPro/Main` yet
- [x] document Tracing by writing a note that explains how to use the library, and the meaning of "reporter", "trace" and "span", as well as the pitfalls
- [x] discuss some of the trade-offs in the implementation, which is why i'm opening this PR already despite it not fully building yet
- [x] it depends on #7789 being merged first
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/7791
GitOrigin-RevId: cadd32d039134c93ddbf364599a2f4dd988adea8
## Description
### I want to speak to the `Manager`
Oh boy. This PR is both fairly straightforward and overreaching, so let's break it down.
For most network access, we need a [`HTTP.Manager`](https://hackage.haskell.org/package/http-client-0.1.0.0/docs/Network-HTTP-Client-Manager.html). It is created only once, at the top level, when starting the engine, and is then threaded through the application to wherever we need to make a network call. As of main, the way we do this is not standardized: most of the GraphQL execution code passes it "manually" as a function argument throughout the code. We also have a custom monad constraint, `HasHttpManagerM`, that describes a monad's ability to provide a manager. And, finally, several parts of the code store the manager in some kind of argument structure, such as `RunT`'s `RunCtx`.
This PR's first goal is to harmonize all of this: we always create the manager at the root, and we already have it when we do our very first `runReaderT`. Wouldn't it make sense for the rest of the code to not manually pass it anywhere, to not store it anywhere, but to always rely on the current monad providing it? This is, in short, what this PR does: it implements a constraint on the base monads, so that they provide the manager, and removes most explicit passing from the code.
### First come, first served
One way this PR goes a tiny bit further than "just" doing the aforementioned harmonization is that it starts the process of implementing the "Services oriented architecture" roughly outlined in this [draft document](https://docs.google.com/document/d/1FAigqrST0juU1WcT4HIxJxe1iEBwTuBZodTaeUvsKqQ/edit?usp=sharing). Instead of using the existing `HasHTTPManagerM`, this PR revamps it into the `ProvidesNetwork` service.
The idea is, again, that we should make all "external" dependencies of the engine, all things that the core of the engine doesn't care about, a "service". This allows us to define clear APIs for features, to choose different implementations based on which version of the engine we're running, harmonizes our many scattered monadic constraints... Which is why this service is called "Network": we can refine it, moving forward, to be the constraint that defines how all network communication is to operate, instead of relying on disparate classes constraint or hardcoded decisions. A comment in the code clarifies this intent.
### Side-effects? In my Haskell?
This PR also unavoidably touches some other aspects of the codebase. One such example: it introduces `Hasura.App.AppContext`, named after `HasuraPro.Context.AppContext`: a name for the reader structure at the base level. It also transforms `Handler` from a type alias to a newtype, as `Handler` is where we actually enforce HTTP limits; but without `Handler` being a distinct type, any code path could simply do a `runExceptT $ runReader` and forget to enforce them.
(As a rule of thumb, i am starting to consider any straggling `runReaderT` or `runExceptT` as a code smell: we should not stack / unstack monads haphazardly, and every layer should be an opaque `newtype` with a corresponding run function.)
## Further work
In several places, i have left TODOs when i have encountered things that suggest that we should do further unrelated cleanups. I'll write down the follow-up steps, either in the aforementioned document or on slack. But, in short, at a glance, in approximate order, we could:
- delete `ExecutionCtx` as it is only a subset of `ServerCtx`, and remove one more `runReaderT` call
- delete `ServerConfigCtx` as it is only a subset of `ServerCtx`, and remove it from `RunCtx`
- remove `ServerCtx` from `HandlerCtx`, and make it part of `AppContext`, or even make it the `AppContext` altogether (since, at least for the OSS version, `AppContext` is there again only a subset)
- remove `CacheBuildParams` and `CacheBuild` altogether, as they're just a distinct stack that is a `ReaderT` on top of `IO` that contains, you guessed it, the same thing as `ServerCtx`
- move `RunT` out of `RQL.Types` and rename it, since after the previous cleanups **it only contains `UserInfo`**; it could be bundled with the authentication service, made a small implementation detail in `Hasura.Server.Auth`
- rename `PGMetadaStorageT` to something a bit more accurate, such as `App`, and enforce its IO base
This would significantly simply our complex stack. From there, or in parallel, we can start moving existing dependencies as Services. For the purpose of supporting read replicas entitlement, we could move `MonadResolveSource` to a `SourceResolver` service, as attempted in #7653, and transform `UserAuthenticationM` into a `Authentication` service.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/7736
GitOrigin-RevId: 68cce710eb9e7d752bda1ba0c49541d24df8209f
Dependencies seem to get concatenated very often, so let's use a data structure that supports efficient concatenation.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/7050
GitOrigin-RevId: 6331963f99f17d1b908a6038318d8c4834cf4dd7
Mostly trying to avoid tricky `Arrows` syntax, and unnecessary use of the `Hasura.Incremental` framework.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6997
GitOrigin-RevId: 9a2f5883e7e29af164e1581049ae003afec2cbe4
What is the `Cacheable` type class about?
```haskell
class Eq a => Cacheable a where
unchanged :: Accesses -> a -> a -> Bool
default unchanged :: (Generic a, GCacheable (Rep a)) => Accesses -> a -> a -> Bool
unchanged accesses a b = gunchanged (from a) (from b) accesses
```
Its only method is an alternative to `(==)`. The added value of `unchanged` (and the additional `Accesses` argument) arises _only_ for one type, namely `Dependency`. Indeed, the `Cacheable (Dependency a)` instance is non-trivial, whereas every other `Cacheable` instance is completely boilerplate (and indeed either generated from `Generic`, or simply `unchanged _ = (==)`). The `Cacheable (Dependency a)` instance is the only one where the `Accesses` argument is not just passed onwards.
The only callsite of the `unchanged` method is in the `ArrowCache (Rule m)` method. That is to say that the `Cacheable` type class is used to decide when we can re-use parts of the schema cache between Metadata operations.
So what is the `Cacheable (Dependency a)` instance about? Normally, the output of a `Rule m a b` is re-used when the new input (of type `a`) is equal to the old one. But sometimes, that's too coarse: it might be that a certain `Rule m a b` only depends on a small part of its input of type `a`. A `Dependency` allows us to spell out what parts of `a` are being depended on, and these parts are recorded as values of types `Access a` in the state `Accesses`.
If the input `a` changes, but not in a way that touches the recorded `Accesses`, then the output `b` of that rule can be re-used without recomputing.
So now you understand _why_ we're passing `Accesses` to the `unchanged` method: `unchanged` is an equality check in disguise that just needs some additional context.
But we don't need to pass `Accesses` as a function argument. We can use the `reflection` package to pass it as type-level context. So the core of this PR is that we change the instance declaration from
```haskell
instance (Cacheable a) => Cacheable (Dependency a) where
```
to
```haskell
instance (Given Accesses, Eq a) => Eq (Dependency a) where
```
and use `(==)` instead of `unchanged`.
If you haven't seen `reflection` before: it's like a `MonadReader`, but it doesn't require a `Monad`.
In order to pass the current `Accesses` value, instead of simply passing the `Accesses` as a function argument, we need to instantiate the `Given Accesses` context. We use the `give` method from the `reflection` package for that.
```haskell
give :: forall r. Accesses -> (Given Accesses => r) -> r
unchanged :: (Given Accesses => Eq a) => Accesses -> a -> a -> Bool
unchanged accesses a b = give accesses (a == b)
```
With these three components in place, we can delete the `Cacheable` type class entirely.
The remainder of this PR is just to remove the `Cacheable` type class and its instances.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6877
GitOrigin-RevId: 7125f5e11d856e7672ab810a23d5bf5ad176e77f
- Avoid a few banana brackets `(| ... |)`, often by just using local `let` bindings
- Use proper `Arrows` syntax rather than helpers like `>->`
- Use monadic `do` syntax instead of `Arrows` syntax where possible
- Avoid `traverseA @Maybe`, in favor of a `case`
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6751
GitOrigin-RevId: c07b22a1a259db6d135486ec71a716705e280717
`CollectedInfo` was just an awkward sum type. By using an explicit `Either` instead, we can guarantee at the type level that certain methods only write inconsistencies, or only write dependencies. This is useful, because if we can guarantee that no dependencies are written, then we don't need to run `resolveDependencies` on that part of the Metadata. In other words, we can keep it out of `BuildOutputs`, which greatly benefits performance - see e.g. hasura/graphql-engine-mono#6613.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6765
GitOrigin-RevId: 9ce099d2eee2278dbb6e5bea72063e4b6e064b35
The main aim of the PR is:
1. To set up a module structure for 'remote-schemas' package.
2. Move parts by the remote schema codebase into the new module structure to validate it.
## Notes to the reviewer
Why a PR with large-ish diff?
1. We've been making progress on the MM project but we don't yet know long it is going to take us to get to the first milestone. To understand this better, we need to figure out the unknowns as soon as possible. Hence I've taken a stab at the first two items in the [end-state](https://gist.github.com/0x777/ca2bdc4284d21c3eec153b51dea255c9) document to figure out the unknowns. Unsurprisingly, there are a bunch of issues that we haven't discussed earlier. These are documented in the 'open questions' section.
1. The diff is large but that is only code moved around and I've added a section that documents how things are moved. In addition, there are fair number of PR comments to help with the review process.
## Changes in the PR
### Module structure
Sets up the module structure as follows:
```
Hasura/
RemoteSchema/
Metadata/
Types.hs
SchemaCache/
Types.hs
Permission.hs
RemoteRelationship.hs
Build.hs
MetadataAPI/
Types.hs
Execute.hs
```
### 1. Types representing metadata are moved
Types that capture metadata information (currently scattered across several RQL modules) are moved into `Hasura.RemoteSchema.Metadata.Types`.
- This new module only depends on very 'core' modules such as
`Hasura.Session` for the notion of roles and `Hasura.Incremental` for `Cacheable` typeclass.
- The requirement on database modules is avoided by generalizing the remote schemas metadata to accept an arbitrary 'r' for a remote relationship
definition.
### 2. SchemaCache related types and build logic have been moved
Types that represent remote schemas information in SchemaCache are moved into `Hasura.RemoteSchema.SchemaCache.Types`.
Similar to `H.RS.Metadata.Types`, this module depends on 'core' modules except for `Hasura.GraphQL.Parser.Variable`. It has something to do with remote relationships but I haven't spent time looking into it. The validation of 'remote relationships to remote schema' is also something that needs to be looked at.
Rips out the logic that builds remote schema's SchemaCache information from the monolithic `buildSchemaCacheRule` and moves it into `Hasura.RemoteSchema.SchemaCache.Build`. Further, the `.SchemaCache.Permission` and `.SchemaCache.RemoteRelationship` have been created from existing modules that capture schema cache building logic for those two components.
This was a fair amount of work. On main, currently remote schema's SchemaCache information is built in two phases - in the first phase, 'permissions' and 'remote relationships' are ignored and in the second phase they are filled in.
While remote relationships can only be resolved after partially resolving sources and other remote schemas, the same isn't true for permissions. Further, most of the work that is done to resolve remote relationships can be moved to the first phase so that the second phase can be a very simple traversal.
This is the approach that was taken - resolve permissions and as much as remote relationships information in the first phase.
### 3. Metadata APIs related types and build logic have been moved
The types that represent remote schema related metadata APIs and the execution logic have been moved to `Hasura.RemoteSchema.MetadataAPI.Types` and `.Execute` modules respectively.
## Open questions:
1. `Hasura.RemoteSchema.Metadata.Types` is so called because I was hoping that all of the metadata related APIs of remote schema can be brought in at `Hasura.RemoteSchema.Metadata.API`. However, as metadata APIs depended on functions from `SchemaCache` module (see [1](ceba6d6226/server/src-lib/Hasura/RQL/DDL/RemoteSchema.hs (L55)) and [2](ceba6d6226/server/src-lib/Hasura/RQL/DDL/RemoteSchema.hs (L91)), it made more sense to create a separate top-level module for `MetadataAPI`s.
Maybe we can just have `Hasura.RemoteSchema.Metadata` and get rid of the extra nesting or have `Hasura.RemoteSchema.Metadata.{Core,Permission,RemoteRelationship}` if we want to break them down further.
1. `buildRemoteSchemas` in `H.RS.SchemaCache.Build` has the following type:
```haskell
buildRemoteSchemas ::
( ArrowChoice arr,
Inc.ArrowDistribute arr,
ArrowWriter (Seq CollectedInfo) arr,
Inc.ArrowCache m arr,
MonadIO m,
HasHttpManagerM m,
Inc.Cacheable remoteRelationshipDefinition,
ToJSON remoteRelationshipDefinition,
MonadError QErr m
) =>
Env.Environment ->
( (Inc.Dependency (HashMap RemoteSchemaName Inc.InvalidationKey), OrderedRoles),
[RemoteSchemaMetadataG remoteRelationshipDefinition]
)
`arr` HashMap RemoteSchemaName (PartiallyResolvedRemoteSchemaCtxG remoteRelationshipDefinition, MetadataObject)
```
Note the dependence on `CollectedInfo` which is defined as
```haskell
data CollectedInfo
= CIInconsistency InconsistentMetadata
| CIDependency
MetadataObject
-- ^ for error reporting on missing dependencies
SchemaObjId
SchemaDependency
deriving (Eq)
```
this pretty much means that remote schemas is dependent on types from databases, actions, ....
How do we fix this? Maybe introduce a typeclass such as `ArrowCollectRemoteSchemaDependencies` which is defined in `Hasura.RemoteSchema` and then implemented in graphql-engine?
1. The dependency on `buildSchemaCacheFor` in `.MetadataAPI.Execute` which has the following signature:
```haskell
buildSchemaCacheFor ::
(QErrM m, CacheRWM m, MetadataM m) =>
MetadataObjId ->
MetadataModifier ->
```
This can be easily resolved if we restrict what the metadata APIs are allowed to do. Currently, they operate in an unfettered access to modify SchemaCache (the `CacheRWM` constraint):
```haskell
runAddRemoteSchema ::
( QErrM m,
CacheRWM m,
MonadIO m,
HasHttpManagerM m,
MetadataM m,
Tracing.MonadTrace m
) =>
Env.Environment ->
AddRemoteSchemaQuery ->
m EncJSON
```
This should instead be changed to restrict remote schema APIs to only modify remote schema metadata (but has access to the remote schemas part of the schema cache), this dependency is completely removed.
```haskell
runAddRemoteSchema ::
( QErrM m,
MonadIO m,
HasHttpManagerM m,
MonadReader RemoteSchemasSchemaCache m,
MonadState RemoteSchemaMetadata m,
Tracing.MonadTrace m
) =>
Env.Environment ->
AddRemoteSchemaQuery ->
m RemoteSchemeMetadataObjId
```
The idea is that the core graphql-engine would call these functions and then call
`buildSchemaCacheFor`.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6291
GitOrigin-RevId: 51357148c6404afe70219afa71bd1d59bdf4ffc6
