### Description
This PR extends the `RemoteSchema` parsers to also include remote relationships. This include a significant refactoring of the top level schema building blocks, since remote schemas can no longer be built in isolation: they have to be built within the same run of `MonadSchema` as the sources. It is originally taken from the changes in #3069 and was slightly adapted.
I highly recommend turning OFF whitespace in the Github UI for `Schema.hs`, since I've adjusted the indentation of two large functions.
### Warning
Given the lack of a feature flag, this PR technically **enables the feature**. While the metadata API is not plugged in, a savvy user could use `replace_metadata` to set a metadata that contains remote joins from remote schemas, and they would be enabled. Is this acceptable?
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3811
GitOrigin-RevId: a5b00f865cdb8890b0fc02b139c2ebd48929f138
### Description
This PR moves Hasura-specific schema functions from `Hasura.GraphQL.Parser.Class` into `Hasura.GraphQL.Schema.Common`. It also removes the two corresponding monad aliases, and consequently harmonizes several parts of the code to use the same common constraint.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3947
GitOrigin-RevId: 40985a7d86da97a311bd480f9a57cc18c350c2a8
### Description
This PR adds the ability to perform remote joins from remote schemas in the engine. To do so, we alter the definition of an `ExecutionStep` targeting a remote schema: the `ExecStepRemote` constructor now expects a `Maybe RemoteJoins`. This new argument is used when processing the execution step, in the transport layer (either `Transport.HTTP` or `Transport.WebSocket`).
For this `Maybe RemoteJoins` to be extracted from a parsed query, this PR also extends the `Execute.RemoteJoin.Collect` module, to implement "collection" from a selection set. Not only do those new functions extract the remote joins, but they also apply all necessary transformations to the selection sets (such as inserting the necessary "phantom" fields used as join keys).
Finally in `Execute.RemoteJoin.Join`, we make two changes. First, we now always look for nested remote joins, regardless of whether the join we just performed went to a source or a remote schema; and second we adapt our join tree logic according to the special cases that were added to deal with remote server edge cases.
Additionally, this PR refactors / cleans / documents `Execute.RemoteJoin.RemoteServer`. This is not required as part of this change and could be moved to a separate PR if needed (a similar cleanup of `Join` is done independently in #3894). It also introduces a draft of a new documentation page for this project, that will be refined in the release PR that ships the feature (either #3069 or a copy of it).
While this PR extends the engine, it doesn't plug such relationships in the schema, meaning that, as of this PR, the new code paths in `Join` are technically unreachable. Adding the corresponding schema code and, ultimately, enabling the metadata API will be done in subsequent PRs.
### Keeping track of concrete type names
The main change this PR makes to the existing `Join` code is to handle a new reserved field we sometimes use when targeting remote servers: the `__hasura_internal_typename` field. In short, a GraphQL selection set can sometimes "branch" based on the concrete "runtime type" of the object on which the selection happens:
```graphql
query {
author(id: 53478) {
... on Writer {
name
articles {
title
}
}
... on Artist {
name
articles {
title
}
}
}
}
```
If both of those `articles` are remote joins, we need to be able, when we get the answer, to differentiate between the two different cases. We do this by asking for `__typename`, to be able to decide if we're in the `Writer` or the `Artist` branch of the query.
To avoid further processing / customization of results, we only insert this `__hasura_internal_typename: __typename` field in the query in the case of unions of interfaces AND if we have the guarantee that we will processing the request as part of the remote joins "folding": that is, if there's any remote join in this branch in the tree. Otherwise, we don't insert the field, and we leave that part of the response untouched.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3810
GitOrigin-RevId: 89aaf16274d68e26ad3730b80c2d2fdc2896b96c
…rmance
It makes sense to try to utilize multiple threads for metadata
operations since we expect them to come one at a time (and likely at
lower load periods anyway).
As noted, although we build roles in parallel now, the admin role is
still a bottleneck. For replace_metadata on huge_schema, on my machine
I get:
BEFORE: 22.7 sec
AFTER: 13.5 sec
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3911
GitOrigin-RevId: 4d4ee6ac8b5506603e70e4fc666a3aacc054d493
### Description
This PR is one further step towards remote joins from remote schemas. It introduces a custom partial AST to represent queries to remote schemas in the IR: we now need to augment what used to be a straightforward GraphQL AST with additional information for remote join fields.
This PR does the minimal amount of work to adjust the rest of the code accordingly, using `Void` in all places that expect a type representing remote relationships.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3794
GitOrigin-RevId: 33fc317731aace71f82ad158a1951ea93350d6cc
In order to respond to GraphQL queries that make use of the introspection fields `__type` or `__schema`, we need two things:
- an overview of the relevant GraphQL type information, stored in a `Schema` object, and
- to have included the `__type` and `__schema` fields in the `query_root` that we generate.
It used to be necessary to do the above items in that order, since the `__type` and `__schema` fields (i.e. the respective `FieldParser`s) were generated _from_ a `Schema` object.
Thanks to recent refactorings in `Hasura.GraphQL.Schema.Introspect` (see hasura/graphql-engine-mono#2835 or hasura/graphql-engine@5760d9289c), the introspection fields _themselves_ are now `Schema`-agnostic, and simply return a function that takes a `Schema` object after parsing. For instance, the type of `schema`, corresponding to the `__schema` field, has literally changed as follows:
```diff
-schema :: MonadParse n => Schema -> FieldParser n ( J.Value)
+schema :: MonadParse n => FieldParser n (Schema -> J.Value)
```
This means that the introspection fields can be included in the GraphQL schema *before* we have generated a `Schema` object. In particular, rather than the current architecture of generating `Schema` at startup time for every role, we can instead generate `Schema` ad-hoc at query parsing time, only for those queries that make use of the introspection fields. This avoids us storing a `Schema` for every role for the lifetime of the server.
However: this introduces a functional change, as the code that generates the `Schema` object, and in particular the `accumulateTypeDefinitions` method, also does certain correctness checks, to prevent exposing a spec-incompliant GraphQL schema. If these correctness checks are being done at parsing time rather than startup time, then we catch certain errors only later on. For this reason, this PR adds an explicit run of this type accumulation at startup time. For efficiency reasons, and since this correctness check is not essential for correct operation of HGE, this is done for the admin role only.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/3231
GitOrigin-RevId: 23701c548b785929b28667025436b6ce60bfe1cd
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
### Description
The GraphQL spec has to conflicting requirements:
1. an object must contain at least one field: the schema may not contain empty objects
2. the _query_root_ must always be present
Given _1_, the schema generation code removes from the schema all fields that would result in empty objects, such as a table for which a user does not have select permissions. But, as a result, our code also potentially removes _query_root_ if it is empty, breaking _2_.
This PR introduces a dummy "placeholder" field in the query root if it's empty, to ensure we never remove it from the schema.
### Remaining work
- [x] changelog entry
- [x] tests
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/148
GitOrigin-RevId: bfd6bfcc2f3de92900b6ba566012f093399ca037
this pr modifies the representation chosen for introspection parsers, "pushing down" the `Schema` input so it is not required to build the parser anymore. instead, the value produced when the parser is evaluated becomes a function that consumes a schema:
```diff
-schema :: MonadParse n => Schema -> FieldParser n ( J.Value)
+schema :: MonadParse n => FieldParser n (Schema -> J.Value)
```
this addresses points (1) and (2) of #2833 and is intended to make #2799 easier: we will need to enforce permissions when generating introspection objects, hiding fields the user is not allowed to see, so if we can pass the schema _later_, we can build this parser once, evaluate it once to (morally) obtain a function `Schema -> Value`, and simply run that single `Schema -> Value` function on different role-based schemas.
(we really need some terminology to be fixed here: "parser" is already not the best name, and then we have parser vs value/function "returned" by parser vs...)
however, we have immediate benefits: we no longer _need_ a `Schema` object to build the introspection parsers! this means we can remove the bogus "degenerate case" schema that is currently constructed in `emptyIntrospection` (and indeed we remove that binding altogether).
(fun fact: the diff for this pull request has a negative line count despite adding a lot of comments. @abooij says i have bragging rights in perpetuity now, à la @nicuveo)
changes:
- internal changes to the operation of the server, invisible outside of a small number of `GraphQL.Schema.*` modules
- no user-facing changes
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2835
Co-authored-by: Auke Booij <164426+abooij@users.noreply.github.com>
Co-authored-by: Brandon Simmons <210815+jberryman@users.noreply.github.com>
GitOrigin-RevId: 9990f53b8f5c733424c4d71a24d94c13dee842ba
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
Source typename customization (hasura/graphql-engine@aac64f2c81) introduced a mechanism to change certain names in the GraphQL schema that is exposed. In particular it allows last-minute modification of:
1. the names of some types, and
2. the names of some root fields.
The above two items are assigned distinct customization algorithms, and at times both algorithms are in scope. So a need to distinguish them is needed.
In the original design, this was addressed by introducing a newtype wrapper `Typename` around GraphQL `Name`s, dedicated to the names of types. However, in the majority of the codebase, type names are also represented by `Name`. For this reason, it was unavoidable to allow for easy conversion. This was supported by a `HasName Typename` instance, as well as by publishing the constructors of `Typename`.
This means that the type safety that newtypes can add is lost. In particular, it is now very easy to confuse type name customization with root field name customization.
This refactors the above design by instead introducing newtypes around the customization operations:
```haskell
newtype MkTypename = MkTypename {runMkTypename :: Name -> Name}
deriving (Semigroup, Monoid) via (Endo Name)
newtype MkRootFieldName = MkRootFieldName {runMkRootFieldName :: Name -> Name}
deriving (Semigroup, Monoid) via (Endo Name)
```
The `Monoid` instance allows easy composition of customization operations, piggybacking off of the type of `Endo`maps.
This design allows safe co-existence of the two customization algorithms, while avoiding the syntactic overhead of packing and unpacking newtypes.
PR-URL: https://github.com/hasura/graphql-engine-mono/pull/2989
GitOrigin-RevId: da3a353a9b003ee40c8d0a1e02872e99d2edd3ca
This commit applies ormolu to the whole Haskell code base by running `make format`.
For in-flight branches, simply merging changes from `main` will result in merge conflicts.
To avoid this, update your branch using the following instructions. Replace `<format-commit>`
by the hash of *this* commit.
$ git checkout my-feature-branch
$ git merge <format-commit>^ # and resolve conflicts normally
$ make format
$ git commit -a -m "reformat with ormolu"
$ git merge -s ours post-ormolu
https://github.com/hasura/graphql-engine-mono/pull/2404
GitOrigin-RevId: 75049f5c12f430c615eafb4c6b8e83e371e01c8e
### Description
We always build a subscription root, even when there was no possible fields. This breaks some third party clients, as the spec does not allow empty types in the schema. This PR fixes this by changing the `buildSubscriptionParser` helper to return a `Maybe` value, and harmonizes / cleans places where we build the subscription root.
https://github.com/hasura/graphql-engine-mono/pull/2357
GitOrigin-RevId: 1aeae25e321eee957e7645c436d17e69207309fd
GJ IR changes cherry-picked from the original GJ branch. There is a separate (can be merged independently) PR for metadata changes (#1727) and there will be a different PR upcoming PR for execution changes.
https://github.com/hasura/graphql-engine-mono/pull/1810
Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com>
GitOrigin-RevId: c31956af29dc9c9b75d002aba7d93c230697c5f4
### Description
This PR adds the required IR for DB to DB joins, based on @paf31 and @0x777 's `feature/db-to-db` branch.
To do so, it also refactors the IR to introduce a new type parameter, `r`, which is used to recursively constructs the `v` parameter of remote QueryDBs. When collecting remote joins, we replace `r` with `Const Void`, indicating at the type level that there cannot be any leftover remote join.
Furthermore, this PR refactors IR.Select for readability, moves some code from IR.Root to IR.Select to avoid having to deal with circular dependencies, and makes it compile by adding `error` in all new cases in the execution pipeline.
The diff doesn't make it clear, but most of Select.hs is actually unchanged. Declarations have just been reordered by topic, in the following order:
- type declarations
- instance declarations
- type aliases
- constructor functions
- traverse functions
https://github.com/hasura/graphql-engine-mono/pull/1580
Co-authored-by: Phil Freeman <630306+paf31@users.noreply.github.com>
GitOrigin-RevId: bbdcb4119cec8bb3fc32f1294f91b8dea0728721
Remote relationships are now supported on SQL Server and BigQuery. The major change though is the re-architecture of remote join execution logic. Prior to this PR, each backend is responsible for processing the remote relationships that are part of their AST.
This is not ideal as there is nothing specific about a remote join's execution that ties it to a backend. The only backend specific part is whether or not the specification of the remote relationship is valid (i.e, we'll need to validate whether the scalars are compatible).
The approach now changes to this:
1. Before delegating the AST to the backend, we traverse the AST, collect all the remote joins while modifying the AST to add necessary join fields where needed.
1. Once the remote joins are collected from the AST, the database call is made to fetch the response. The necessary data for the remote join(s) is collected from the database's response and one or more remote schema calls are constructed as necessary.
1. The remote schema calls are then executed and the data from the database and from the remote schemas is joined to produce the final response.
### Known issues
1. Ideally the traversal of the IR to collect remote joins should return an AST which does not include remote join fields. This operation can be type safe but isn't taken up as part of the PR.
1. There is a lot of code duplication between `Transport/HTTP.hs` and `Transport/Websocket.hs` which needs to be fixed ASAP. This too hasn't been taken up by this PR.
1. The type which represents the execution plan is only modified to handle our current remote joins and as such it will have to be changed to accommodate general remote joins.
1. Use of lenses would have reduced the boilerplate code to collect remote joins from the base AST.
1. The current remote join logic assumes that the join columns of a remote relationship appear with their names in the database response. This however is incorrect as they could be aliased. This can be taken up by anyone, I've left a comment in the code.
### Notes to the reviewers
I think it is best reviewed commit by commit.
1. The first one is very straight forward.
1. The second one refactors the remote join execution logic but other than moving things around, it doesn't change the user facing functionality. This moves Postgres specific parts to `Backends/Postgres` module from `Execute`. Some IR related code to `Hasura.RQL.IR` module. Simplifies various type class function signatures as a backend doesn't have to handle remote joins anymore
1. The third one fixes partial case matches that for some weird reason weren't shown as warnings before this refactor
1. The fourth one generalizes the validation logic of remote relationships and implements `scalarTypeGraphQLName` function on SQL Server and BigQuery which is used by the validation logic. This enables remote relationships on BigQuery and SQL Server.
https://github.com/hasura/graphql-engine-mono/pull/1497
GitOrigin-RevId: 77dd8eed326602b16e9a8496f52f46d22b795598
This reverts the remote schema type customisation and namespacing feature temporarily as we test for certain conditions.
GitOrigin-RevId: f8ee97233da4597f703970c3998664c03582d8e7
Fixes https://github.com/hasura/graphql-engine-mono/issues/712
Main point of interest: the `Hasura.SQL.Backend` module.
This PR creates an `Exists` type indexed by indexed type and packed constraint while hiding all of its complexity by not exporting the constructor.
Existential constructors/types which are no longer (directly) existential:
- [X] BackendSourceInfo :: BackendSourceInfo
- [x] BackendSourceMetadata :: BackendSourceMetadata
- [x] MOSourceObjId :: MetadatObjId
- [x] SOSourceObj :: SchemaObjId
- [x] RFDB :: RootField
- [x] LQP :: LiveQueryPlan
- [x] ExecutionStep :: ExecStepDB
This PR also removes ALL usages of `Typeable.cast` from our codebase. We still need to derive `Typeable` in a few places in order to be able to derive `Data` in one place. I have not dug deeper to see why this is needed.
GitOrigin-RevId: bb47e957192e4bb0af4c4116aee7bb92f7983445
fixes#3868
docker image - `hasura/graphql-engine:inherited-roles-preview-48b73a2de`
Note:
To be able to use the inherited roles feature, the graphql-engine should be started with the env variable `HASURA_GRAPHQL_EXPERIMENTAL_FEATURES` set to `inherited_roles`.
Introduction
------------
This PR implements the idea of multiple roles as presented in this [paper](https://www.microsoft.com/en-us/research/wp-content/uploads/2016/02/FGALanguageICDE07.pdf). The multiple roles feature in this PR can be used via inherited roles. An inherited role is a role which can be created by combining multiple singular roles. For example, if there are two roles `author` and `editor` configured in the graphql-engine, then we can create a inherited role with the name of `combined_author_editor` role which will combine the select permissions of the `author` and `editor` roles and then make GraphQL queries using the `combined_author_editor`.
How are select permissions of different roles are combined?
------------------------------------------------------------
A select permission includes 5 things:
1. Columns accessible to the role
2. Row selection filter
3. Limit
4. Allow aggregation
5. Scalar computed fields accessible to the role
Suppose there are two roles, `role1` gives access to the `address` column with row filter `P1` and `role2` gives access to both the `address` and the `phone` column with row filter `P2` and we create a new role `combined_roles` which combines `role1` and `role2`.
Let's say the following GraphQL query is queried with the `combined_roles` role.
```graphql
query {
employees {
address
phone
}
}
```
This will translate to the following SQL query:
```sql
select
(case when (P1 or P2) then address else null end) as address,
(case when P2 then phone else null end) as phone
from employee
where (P1 or P2)
```
The other parameters of the select permission will be combined in the following manner:
1. Limit - Minimum of the limits will be the limit of the inherited role
2. Allow aggregations - If any of the role allows aggregation, then the inherited role will allow aggregation
3. Scalar computed fields - same as table column fields, as in the above example
APIs for inherited roles:
----------------------
1. `add_inherited_role`
`add_inherited_role` is the [metadata API](https://hasura.io/docs/1.0/graphql/core/api-reference/index.html#schema-metadata-api) to create a new inherited role. It accepts two arguments
`role_name`: the name of the inherited role to be added (String)
`role_set`: list of roles that need to be combined (Array of Strings)
Example:
```json
{
"type": "add_inherited_role",
"args": {
"role_name":"combined_user",
"role_set":[
"user",
"user1"
]
}
}
```
After adding the inherited role, the inherited role can be used like single roles like earlier
Note:
An inherited role can only be created with non-inherited/singular roles.
2. `drop_inherited_role`
The `drop_inherited_role` API accepts the name of the inherited role and drops it from the metadata. It accepts a single argument:
`role_name`: name of the inherited role to be dropped
Example:
```json
{
"type": "drop_inherited_role",
"args": {
"role_name":"combined_user"
}
}
```
Metadata
---------
The derived roles metadata will be included under the `experimental_features` key while exporting the metadata.
```json
{
"experimental_features": {
"derived_roles": [
{
"role_name": "manager_is_employee_too",
"role_set": [
"employee",
"manager"
]
}
]
}
}
```
Scope
------
Only postgres queries and subscriptions are supported in this PR.
Important points:
-----------------
1. All columns exposed to an inherited role will be marked as `nullable`, this is done so that cell value nullification can be done.
TODOs
-------
- [ ] Tests
- [ ] Test a GraphQL query running with a inherited role without enabling inherited roles in experimental features
- [] Tests for aggregate queries, limit, computed fields, functions, subscriptions (?)
- [ ] Introspection test with a inherited role (nullability changes in a inherited role)
- [ ] Docs
- [ ] Changelog
Co-authored-by: Vamshi Surabhi <6562944+0x777@users.noreply.github.com>
GitOrigin-RevId: 3b8ee1e11f5ceca80fe294f8c074d42fbccfec63
