## Description
This change adds support for querying into nested arrays in Data Connector agents that support such a concept (currently MongoDB).
### DC API changes
- New API type `ColumnType` which allows representing the type of a "column" as either a scalar type, an object reference or an array of `ColumnType`s. This recursive definition allows arbitrary nesting of arrays of types.
- The `type` fields in the API types `ColumnInfo` and `ColumnInsertSchema` now take a `ColumnType` instead of a `ScalarType`.
- To ensure backwards compatibility, a `ColumnType` representing a scalar serialises and deserialises to the same representation as `ScalarType`.
- In queries, the `Field` type now has a new constructor `NestedArrayField`. This contains a nested `Field` along with optional `limit`, `offset`, `where` and `order_by` arguments. (These optional arguments are not yet used by either HGE or the MongoDB agent.)
### MongoDB Haskell agent changes
- The `/schema` endpoint will now recognise arrays within the JSON validation schema and generate corresponding arrays in the DC schema.
- The `/query` endpoint will now handle `NestedArrayField`s within queries (although it does not yet handle `limit`, `offset`, `where` and `order_by`).
### HGE server changes
- The `Backend` type class adds a new type family `XNestedArrays b` to enable nested arrays on a per-backend basis (currently enabled only for the `DataConnector` backend.
- Within `RawColumnInfo` the column type is now represented by a new type `RawColumnType b` which mirrors the shape of the DC API `ColumnType`, but uses `XNestedObjects b` and `XNestedArrays b` type families to allow turning nested object and array supports on or off for a particular backend. In the `DataConnector` backend `API.CustomType` is converted into `RawColumnInfo 'DataConnector` while building the schema.
- In the next stage of schema building, the `RawColumnInfo` is converted into a `StructuredColumnInfo` which allows us to represent the three different types of columns: scalar, object and array. TODO: the `StructuredColumnInfo` looks very similar to the Logical Model types. The main difference is that it uses the `XNestedObjects` and `XNestedArrays` type families. We should be able to combine these two representations.
- The `StructuredColumnInfo` is then placed into a `FIColumn` `FieldInfo`. This involved some refactoring of `FieldInfo` as I had previously split out `FINestedObject` into a separate constructor. However it works out better to represent all "column" fields (i.e. scalar, object and array) using `FIColumn` as this make it easier to implement permission checking correctly. This is the reason the `StructuredColumnInfo` was needed.
- Next, the `FieldInfo` are used to generate `FieldParser`s. We add a new constructor to `AnnFieldG` for `AFNestedArray`. An `AFNestedArray` field parser can contain either a simple array selection or an array aggregate. Simple array `FieldParsers` are currently limited to subfield selection. We will add support for limit, offset, where and order_by in a future PR. We also don't yet generate array aggregate `FieldParsers.
- The new `AFNestedArray` field is handled by the `QueryPlan` module in the `DataConnector` backend. There we generate an `API.NestedArrayField` from the AFNestedArray. We also handle nested arrays when reshaping the response from the DC agent.
## Limitations
- Support for limit, offset, filter (where) and order_by is not yet fully implemented, although it should not be hard to add this
- Support for aggregations on nested arrays is not yet fully implemented
- Permissions involving nested arrays (and objects) not yet implemented
- This should be integrated with Logical Model types, but that will happen in a separate PR
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/9149
GitOrigin-RevId: 0e7b71a994fc1d2ca1ef73bfe7b96e95b5328531
## Description
This change adds support for nested object fields in HGE IR and Schema Cache, the Data Connectors backend and API, and the MongoDB agent.
### Data Connector API changes
- The `/schema` endpoint response now includes an optional set of GraphQL type definitions. Table column types can refer to these definitions by name.
- Queries can now include a new field type `object` which contains a column name and a nested query. This allows querying into a nested object within a field.
### MongoDB agent changes
- Add support for querying into nested documents using the new `object` field type.
### HGE changes
- The `Backend` type class has a new type family `XNestedObjects b` which controls whether or not a backend supports querying into nested objects. This is currently enabled only for the `DataConnector` backend.
- For backends that support nested objects, the `FieldInfo` type gets a new constructor `FINestedObject`, and the `AnnFieldG` type gets a new constructor `AFNestedObject`.
- If the DC `/schema` endpoint returns any custom GraphQL type definitions they are stored in the `TableInfo` for each table in the source.
- During schema cache building, the function `addNonColumnFields` will check whether any column types match custom GraphQL object types stored in the `TableInfo`. If so, they are converted into `FINestedObject` instead of `FIColumn` in the `FieldInfoMap`.
- When building the `FieldParser`s from `FieldInfo` (function `fieldSelection`) any `FINestedObject` fields are converted into nested object parsers returning `AFNestedObject`.
- The `DataConnector` query planner converts `AFNestedObject` fields into `object` field types in the query sent to the agent.
## Limitations
### HGE not yet implemented:
- Support for nested arrays
- Support for nested objects/arrays in mutations
- Support for nested objects/arrays in order-by
- Support for filters (`where`) in nested objects/arrays
- Support for adding custom GraphQL types via track table metadata API
- Support for interface and union types
- Tests for nested objects
### Mongo agent not yet implemented:
- Generate nested object types from validation schema
- Support for aggregates
- Support for order-by
- Configure agent port
- Build agent in CI
- Agent tests for nested objects and MongoDB agent
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/7844
GitOrigin-RevId: aec9ec1e4216293286a68f9b1af6f3f5317db423
### Description
This monster of a PR took way too long. As the title suggests, it reduces the schema context carried in the readers to the very strict minimum. In practice, that means that to build a source, we only require:
- the global `SchemaContext`
- the global `SchemaOptions` (soon to be renamed `SchemaSourceOptions`)
- that source's `SourceInfo`
Furthermore, _we no longer carry "default" customization options throughout the schema_. All customization information is extracted from the `SourceInfo`, when required. This prevents an entire category of bugs we had previously encountered, such as parts of the code using uninitialized / unupdated customization info.
In turn, this meant that we could remove the explicit threading of the `SourceInfo` throughout the schema, since it is now always available through the reader context.
Finally, this meant making a few adjustments to relay and actions as well, such as the introduction of a new separate "context" for actions, and a change to how we create some of the action-specific postgres scalar parsers.
I'll highlight with review comments the areas of interest.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/6709
GitOrigin-RevId: ea80fddcb24e2513779dd04b0b700a55f0028dd1
### Description
This PR changes all the schema code to operate in a specific `SchemaT` monad, rather than in an arbitrary `m` monad. `SchemaT` is intended to be used opaquely with `runSourceSchema` and `runRemoteSchema`. The main goal of this is to allow a different reader context per part of the schema: this PR also minimizes the contexts. This means that we no longer require `SchemaOptions` when building remote schemas' schema, and this PR therefore removes a lot of dummy / placeholder values accordingly.
### Performance and stacking
This PR has been through several iterations. #5339 was the original version, that accomplished the same thing by stacking readers on top of the stack at every remote relationship boundary. This raised performance concerns, and @0x777 confirmed with an ad-hoc test that in some extreme cases we could see up to a 10% performance impact. This version, while more verbose, allows us to unstack / re-stack the readers, and avoid that problem. #5517 adds a new benchmark set to be able to automatically measure this on every PR.
### Remaining work
- [x] a comment (or perhaps even a Note?) should be added to `SchemaT`
- [x] we probably want for #5517 to be merged first so that we can confirm the lack of performance penalty
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5458
GitOrigin-RevId: e06b83d90da475f745b838f1fd8f8b4d9d3f4b10
Followup to hasura/graphql-engine-mono#4713.
The `memoizeOn` method, part of `MonadSchema`, originally had the following type:
```haskell
memoizeOn
:: (HasCallStack, Ord a, Typeable a, Typeable b, Typeable k)
=> TH.Name
-> a
-> m (Parser k n b)
-> m (Parser k n b)
```
The reason for operating on `Parser`s specifically was that the `MonadSchema` effect would additionally initialize certain `Unique` values, which appear (nested in) the type of `Parser`.
hasura/graphql-engine-mono#518 changed the type of `memoizeOn`, to additionally allow memoizing `FieldParser`s. These also contained a `Unique` value, which was similarly initialized by the `MonadSchema` effect. The new type of `memoizeOn` was as follows:
```haskell
memoizeOn
:: forall p d a b
. (HasCallStack, HasDefinition (p n b) d, Ord a, Typeable p, Typeable a, Typeable b)
=> TH.Name
-> a
-> m (p n b)
-> m (p n b)
```
Note the type `p n b` of the value being memoized: by choosing `p` to be either `Parser k` or `FieldParser`, both can be memoized. Also note the new `HasDefinition (p n b) d` constraint, which provided a `Lens` for accessing the `Unique` value to be initialized.
A quick simplification is that the `HasCallStack` constraint has never been used by any code. This was realized in hasura/graphql-engine-mono#4713, by removing that constraint.
hasura/graphql-engine-mono#2980 removed the `Unique` value from our GraphQL-related types entirely, as their original purpose was never truly realized. One part of removing `Unique` consisted of dropping the `HasDefinition (p n b) d` constraint from `memoizeOn`.
What I didn't realize at the time was that this meant that the type of `memoizeOn` could be generalized and simplified much further. This PR finally implements that generalization. The new type is as follows:
```haskell
memoizeOn ::
forall a p.
(Ord a, Typeable a, Typeable p) =>
TH.Name ->
a ->
m p ->
m p
```
This change has a couple of consequences.
1. While constructing the schema, we often output `Maybe (Parser ...)`, to model that the existence of certain pieces of GraphQL schema sometimes depends on the permissions that a certain role has. The previous versions of `memoizeOn` were not able to handle this, as the only thing they could memoize was fully-defined (if not yet fully-evaluated) `(Field)Parser`s. This much more general API _would_ allow memoizing `Maybe (Parser ...)`s. However, we probably have to be continue being cautious with this: if we blindly memoize all `Maybe (Parser ...)`s, the resulting code may never be able to decide whether the value is `Just` or `Nothing` - i.e. it never commits to the existence-or-not of a GraphQL schema fragment. This would manifest as a non-well-founded knot tying, and this would get reported as an error by the implementation of `memoizeOn`.
tl;dr: This generalization _technically_ allows for memoizing `Maybe` values, but we probably still want to avoid doing so.
For this reason, the PR adds a specialized version of `memoizeOn` to `Hasura.GraphQL.Schema.Parser`.
2. There is no longer any need to connect the `MonadSchema` knot-tying effect with the `MonadParse` effect. In fact, after this PR, the `memoizeOn` method is completely GraphQL-agnostic, and so we implement hasura/graphql-engine-mono#4726, separating `memoizeOn` from `MonadParse` entirely - `memoizeOn` can be defined and implemented as a general Haskell typeclass method.
Since `MonadSchema` has been made into a single-type-parameter type class, it has been renamed to something more general, namely `MonadMemoize`. Its only task is to memoize arbitrary `Typeable p` objects under a combined key consisting of a `TH.Name` and a `Typeable a`.
Also for this reason, the new `MonadMemoize` has been moved to the more general `Control.Monad.Memoize`.
3. After this change, it's somewhat clearer what `memoizeOn` does: it memoizes an arbitrary value of a `Typeable` type. The only thing that needs to be understood in its implementation is how the manual blackholing works. There is no more semantic interaction with _any_ GraphQL code.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4725
Co-authored-by: Daniel Harvey <4729125+danieljharvey@users.noreply.github.com>
GitOrigin-RevId: 089fa2e82c2ce29da76850e994eabb1e261f9c92
### Description
I am not 100% sure about this PR; while I think the code is better this way, I'm willing to be convinced otherwise.
In short, this PR moves the `RoleName` field into the `SchemaContext`, instead of being a nebulous `Has RoleName` constraint on the reader monad. The major upside of this is that it makes it an explicit named field, rather than something that must be given as part of a tuple of arguments when calling `runReader`.
However, the downside is that it breaks the helper permissions functions of `Schema.Table`, which relied on `Has RoleName r`. This PR makes the choice of passing the role name explicitly to all of those functions, which in turn means first explicitly fetching the role name in a lot of places. It makes it more explicit when a schema building block relies on the role name, but is a bit verbose...
### Alternatives
Some alternatives worth considering:
- attempting something like `Has context r, Has RoleName context`, which would allow them to be independent from the context but still fetch the role name from the reader, but might require type annotations to not be ambiguous
- keeping the permission functions the same, with `Has RoleName r`, and introducing a bunch of newtypes instead of using tuples to explicitly implement all the required `Has` instances
- changing the permission functions to `Has SchemaContext r`, since they are functions used only to build the schema, and therefore may be allowed to be tied to the context.
What do y'all think?
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/5073
GitOrigin-RevId: 8fd09fafb54905a4d115ef30842d35da0c3db5d2
This moves `MkTypename` and `NamingCase` into their own modules, with the intent of reducing the scope of the schema parsers code, and trying to reduce imports of large modules when small ones will do.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4978
GitOrigin-RevId: 19541257fe010035390f6183a4eaa37bae0d3ca1
The code that builds the GraphQL schema, and `buildGQLContext` in particular, is partial: not every value of `(ServerConfigCtx, GraphQLQueryType, SourceCache, HashMap RemoteSchemaName (RemoteSchemaCtx, MetadataObject), ActionCache, AnnotatedCustomTypes)` results in a valid GraphQL schema. When it fails, we want to be able to return better error messages than we currently do.
The key thing that is missing is a way to trace back GraphQL type information to their origin from the Hasura metadata. Currently, we have a number of correctness checks of our GraphQL schema. But these correctness checks only have access to pure GraphQL type information, and hence can only report errors in terms of that. Possibly the worst is the "conflicting definitions" error, which, in practice, can only be debugged by Hasura engineers. This is terrible DX for customers.
This PR allows us to print better error messages, by adding a field to the `Definition` type that traces the GraphQL type to its origin in the metadata. So the idea is simple: just add `MetadataObjId`, or `Maybe` that, or some other sum type of that, to `Definition`.
However, we want to avoid having to import a `Hasura.RQL` module from `Hasura.GraphQL.Parser`. So we instead define this additional field of `Definition` through a new type parameter, which is threaded through in `Hasura.GraphQL.Parser`. We then define type synonyms in `Hasura.GraphQL.Schema.Parser` that fill in this type parameter, so that it is not visible for the majority of the codebase.
The idea of associating metadata information to `Definition`s really comes to fruition when combined with hasura/graphql-engine-mono#4517. Their combination would allow us to use the API of fatal errors (just like the current `MonadError QErr`) to report _inconsistencies_ in the metadata. Such inconsistencies are then _automatically_ ignored. So no ad-hoc decisions need to be made on how to cut out inconsistent metadata from the GraphQL schema. This will allow us to report much better errors, as well as improve the likelihood of a successful HGE startup.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4770
Co-authored-by: Samir Talwar <47582+SamirTalwar@users.noreply.github.com>
GitOrigin-RevId: 728402b0cae83ae8e83463a826ceeb609001acae
Pretty much all quasi-quoted names in the server code base have ended up in `Hasura.GraphQL.Parser.Constants`. I'm now finding this unpleasant for two reasons:
1. I would like to factor out the parser code into its own Cabal package, and I don't want to have to expose all these names.
2. Most of them really have nothing to do with the parsers.
In order to remedy this, I have:
1. moved the names used by parser code to `Hasura.GraphQL.Parser.DirectiveName`, as they're all related to directives;
2. moved `Hasura.GraphQL.Parser.Constants` to `Hasura.Name`, changing the qualified import name from `G` to `Name`;
3. moved names only used in tests to the appropriate test case;
4. removed unused items from `Hasura.Name`; and
5. grouped related names.
Most of the changes are simply changing `G` to `Name`, which I find much more meaningful.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4777
GitOrigin-RevId: a77aa0aee137b2b5e6faec94495d3a9fbfa1348b
This is a first step towards clarifying the role of `UnpreparedValue` as part of the IR. It certainly does not belong in the parser framework.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4588
GitOrigin-RevId: d1582a0b266729b79e00d31057178a4099168e6d
### Description
The main goal of this PR is, as stated, to remove the circular dependency in the schema building code. This cycle arises from the existence of remote relationships: when we build the schema for a source A, a remote relationship might force us to jump to the schema of a source B, or some remote schema. As a result, we end up having to do a dispatch from a "leaf" of the schema, similar to the one done at the root. In turn, this forces us to carry along in the schema a lot of information required for that dispatch, AND it forces us to import the instances in scope, creating an import loop.
As discussed in #4489, this PR implements the "dependency injection" solution: we pass to the schema a function to call to do the dispatch, and to get a generated field for a remote relationship. That way, this function can be chosen at the root level, and the leaves need not be aware of the overall context.
This PR grew a bit bigger than that, however; in an attempt to try and remove the `SourceCache` from the schema altogether, it changed a lot of functions across the schema building code, to thread along the `SourceInfo b` of the source being built. This avoids having to do cache lookups within a given source. A few cases remain, such as relay, that we might try to tackle in a subsequent PR.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4557
GitOrigin-RevId: 9388e48372877520a72a9fd1677005df9f7b2d72
## Description
This PR removes `RQL.Types`, which was now only re-exporting a bunch of unrelated modules.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4363
GitOrigin-RevId: 894f29a19bff70b3dad8abc5d9858434d5065417
## Description
This small PR moves all functions in `RQL.Types.hs` to better locations. Most `askX` functions are moved alongside the `unsafe` functions they use. Several other functions are moved closer to their call site. `MetadataM` is moved alongside `Metadata`. This PR also documents the `ask` functions.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/4355
GitOrigin-RevId: 0498a7e8f98e7a94af911dd375cad84ace7ddffa
### Description
This PR improves the `Collect` module by re-ordering the functions to make clear what is public API and what is internal implementation. Furthermore, it makes use of `traverseOf` and `traverseFields` to reduce duplication. To do so, it also introduces a few more lenses in the rest of the codebase, and uses this opportunity to harmonize some structures that were not following our naming convention.
While the diff is massive, a lot of it is just code moving around; the file is now divided into separate sections:
- entry points: IR types for which we want to run the collection
- internal monadic structure
- internal traversals: functions that do nothing but drill down further
- actual transformations: the three cases where we do actually have work to do: selection sets on which we do want to insert join columns, extract remote relationships... those functions are left unchanged by this PR
- internal helpers
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3863
GitOrigin-RevId: f7cbecfae9eed9737b62acfa5848bfcf9d4651f6
We build the GraphQL schema by combining building blocks such as `tableSelectionSet` and `columnParser`. These building blocks individually build `{InputFields,Field,}Parser` objects. Those object specify the valid GraphQL schema.
Since the GraphQL schema is role-dependent, at some point we need to know what fragment of the GraphQL schema a specific role is allowed to access, and this is stored in `{Sel,Upd,Ins,Del}PermInfo` objects.
We have passed around these permission objects as function arguments to the schema building blocks since we first started dealing with permissions during the PDV refactor - see hasura/graphql-engine@5168b99e46 in hasura/graphql-engine#4111. This means that, for instance, `tableSelectionSet` has as its type:
```haskell
tableSelectionSet ::
forall b r m n.
MonadBuildSchema b r m n =>
SourceName ->
TableInfo b ->
SelPermInfo b ->
m (Parser 'Output n (AnnotatedFields b))
```
There are three reasons to change this.
1. We often pass a `Maybe (xPermInfo b)` instead of a proper `xPermInfo b`, and it's not clear what the intended semantics of this is. Some potential improvements on the data types involved are discussed in issue hasura/graphql-engine-mono#3125.
2. In most cases we also already pass a `TableInfo b`, and together with the `MonadRole` that is usually also in scope, this means that we could look up the required permissions regardless: so passing the permissions explicitly undermines the "single source of truth" principle. Breaking this principle also makes the code more difficult to read.
3. We are working towards role-based parsers (see hasura/graphql-engine-mono#2711), where the `{InputFields,Field,}Parser` objects are constructed in a role-invariant way, so that we have a single object that can be used for all roles. In particular, this means that the schema building blocks _need_ to be constructed in a role-invariant way. While this PR doesn't accomplish that, it does reduce the amount of role-specific arguments being passed, thus fixing hasura/graphql-engine-mono#3068.
Concretely, this PR simply drops the `xPermInfo b` argument from almost all schema building blocks. Instead these objects are looked up from the `TableInfo b` as-needed. The resulting code is considerably simpler and shorter.
One way to interpret this change is as follows. Before this PR, we figured out permissions at the top-level in `Hasura.GraphQL.Schema`, passing down the obtained `xPermInfo` objects as required. After this PR, we have a bottom-up approach where the schema building blocks themselves decide whether they want to be included for a particular role.
So this moves some permission logic out of `Hasura.GraphQL.Schema`, which is very complex.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3608
GitOrigin-RevId: 51a744f34ec7d57bc8077667ae7f9cb9c4f6c962
This PR simplifies the types that represent a remote relationship in IR so that they can be reused in other parts (in remote schema types) which could have remote relationships.
The comments on the PR explain the main changes.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2979
GitOrigin-RevId: 559c51d9d6ae79e2183ce4347018741b9096ac74
GraphQL types can refer to each other in a circular way. The PDV framework used to use values of type `Unique` to recognize two fragments of GraphQL schema as being the same instance. Internally, this is based on `Data.Unique` from the `base` package, which simply increases a counter on every creation of a `Unique` object.
**NB**: The `Unique` values are _not_ used for knot tying the schema combinators themselves (i.e. `Parser`s). The knot tying for `Parser`s is purely based on keys provided to `memoizeOn`. The `Unique` values are _only_ used to recognize two pieces of GraphQL _schema_ as being identical. Originally, the idea was that this would help us with a perfectly correct identification of GraphQL types. But this fully correct equality checking of GraphQL types was never implemented, and does not seem to be necessary to prevent bugs.
Specifically, these `Unique` values are stored as part of `data Definition a`, which specifies a part of our internal abstract syntax tree for the GraphQL types that we expose. The `Unique` values get initialized by the `SchemaT` effect.
In #2894 and #2895, we are experimenting with how (parts of) the GraphQL types can be hidden behind certain permission predicates. This would allow a single GraphQL schema in memory to serve all roles, implementing #2711. The permission predicates get evaluated at query parsing time when we know what role is doing a certain request, thus outputting the correct GraphQL types for that role.
If the approach of #2895 is followed, then the `Definition` objects, and thus the `Unique` values, would be hidden behind the permission predicates. Since the permission predicates are evaluated only after the schema is already supposed to be built, this means that the permission predicates would prevent us from initializing the `Unique` values, rendering them useless.
The simplest remedy to this is to remove our usage of `Unique` altogether from the GraphQL schema and schema combinators. It doesn't serve a functional purpose, doesn't prevent bugs, and requires extra bookkeeping.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2980
GitOrigin-RevId: 50d3f9e0b9fbf578ac49c8fc773ba64a94b1f43d
