Introduce C# driver without deployment #609

farost · 2024-03-15T12:16:49Z

Usage and product changes

We introduce the C# driver for TypeDB. It is built using the cross-platform .NET 6 framework.

Usage: Deployment and usage examples will be provided in a separate pull request. Current state of the code lets you compiling the driver + writing and running behaviour and integration tests for it.
The driver is expected to be a Nuget package, which could be added as a dependency to a project and referenced via "using" statements inside the users' code for all platforms.

Architecture: The C# driver is a thin wrapper around the TypeDB Rust driver, introducing classes for a more intuitive interface. Mostly each C# object holds a reference to the corresponding native Rust object, using an FFI (SWIG for C#) for the native object wrappers generation and resource management.

Any error encountered will throw a TypeDBDriverException. Note that methods which return an IEnumerable or a Promise and encounter a server-side error will only throw when the return objects are evaluated (e.g. iterate over or call a Linq method for an IEnumerable and call Resolve() for a Promise).

A simple usage example (more to come after the integration tests are finished in another PR):

// Inside a try-catch block
using (ITypeDBDriver driver = TypeDB.CoreDriver(TypeDB.DEFAULT_ADDRESS))
{
    string dbName = "mydb";
    driver.Databases.Create(dbName);
    IDatabase mydb = driver.Databases.Get(dbName);
    System.Console.WriteLine(mydb.Name);

    using (ITypeDBSession schemaSession = driver.Session(dbName, SessionType.SCHEMA))
    {
        using (ITypeDBTransaction writeTransaction = schemaSession.Transaction(TransactionType.WRITE))
        {
            string defineQuery = "...some define query...";
            writeTransaction.Query.Define(defineQuery).Resolve();
            writeTransaction.Commit();
        }
    }

    mydb.Delete();
}

Implementation

Driver:
Introduces the C# driver, wrapping the C-native classes produced by Rust driver using SWIG. The driver aims to provide a familiar interface for C# developers, hiding data transformations and native calls under the hood.
The code's architecture and build flow follow the one from Java with small C#-specific alterations.

Tests:
Integration tests use NUnit framework as a built-in framework for C# bazel test rules (more to come in another PR, right now these tests are messy).
Behaviour tests use Xunit.Gherkin.Quick framework and are implemented using partial classes for flexible construction of tests based on different sets of step declaration files.

Build:
To fetch nuget packages, we use Paket and reference this dependencies in a regular Bazel way (example from the official Bazel repo for C# rules).
The corresponding dependencies PR for C# rules is here.

typedb-bot · 2024-03-15T12:16:53Z

PR Review Checklist

Do not edit the content of this comment. The PR reviewer should simply update this comment by ticking each review item below, as they get completed.

Trivial Change

This change is trivial and does not require a code or architecture review.

Code

Packages, classes, and methods have a single domain of responsibility.
Packages, classes, and methods are grouped into cohesive and consistent domain model.
The code is canonical and the minimum required to achieve the goal.
Modules, libraries, and APIs are easy to use, robust (foolproof and not errorprone), and tested.
Logic and naming has clear narrative that communicates the accurate intent and responsibility of each module (e.g. method, class, etc.).
The code is algorithmically efficient and scalable for the whole application.

Architecture

Any required refactoring is completed, and the architecture does not introduce technical debt incidentally.
Any required build and release automations are updated and/or implemented.
Any new components follows a consistent style with respect to the pre-existing codebase.
The architecture intuitively reflects the application domain, and is easy to understand.
The architecture has a well-defined hierarchy of encapsulated components.
The architecture is extensible and scalable.

farost · 2024-03-15T13:45:20Z

A quick comment about Naming

A typical class's signature is:

class ClassName
{
    private int _privateField;
    public int PublicField;

    void MethodName(int argumentName)
    {
        ...
    }
}

C# identifier names coding conventions
C# common code conventions

Exception: we decided to mark public constants LIKE_THIS, because the rules from the Microsoft docs were not clear... Maybe we'll come up with something within this review.

Namespaces and directories:
I try to follow what Microsoft says, considering Haikal's recent comment: "Please don't ever spell it Typedb":
C# Names of Namespaces

After viewing some of the open-source C# projects, we came up with naming everything starting with an Uppercase char and declaring one namespace per library (Api, Common, Concept, Connection, ...). The last rule is broken in Behaviour/Utils, but I'll additionally comment it inside the code.

It also means that to run a test like in java:
bazel test //java/test/behaviour/connection/transaction:test-core;
We need to uppercase everything inside the csharp's path:
bazel test //csharp/Test/Behaviour/Connection/Transaction:test-core;

farost

Tried to clean everything up, it's ready for your review.

csharp/Api/Answer/IConceptMap.cs

farost · 2024-03-15T13:55:54Z

c/swig/typedb_driver_csharp.swg

+%}
+
+
+%typemap(csbase) Error "System.Exception";


An example from the SWIG C# doc. This block rethrows native exceptions to C#.

farost · 2024-03-15T13:56:27Z

c/swig/typedb_driver_csharp.swg

+%csconst(1);
+
+%typemap(cscode) SWIGTYPE %{
+ public $csclassname Released()


Release memory in a Java-like way.

farost · 2024-03-15T13:58:01Z

c/swig/typedb_driver_csharp.swg

+ static NativeExceptionHelper exceptionHelper = new NativeExceptionHelper();
+%}
+
+%exception %{


This code is pasted in the generated files after every native call (except of the calls marked %noexception lower, it is the same for Java).

farost · 2024-03-15T13:58:51Z

c/swig/typedb_driver_csharp.swg

+%noexception transaction_commit;
+%noexception transaction_rollback;
+
+%define %promise(Prefix, Type, function_prefix)


Our Promises. Decided to implement it the same way as in java without architectural refactoring.

java/test/behaviour/query/QuerySteps.java

farost · 2024-03-15T16:05:13Z

java/test/behaviour/query/QuerySteps.java

 TypeQLInsert typeQLQuery = TypeQL.parseQuery(String.join("\n", insertQueryStatements));
- return tx().query().insert(String.join("\n", insertQueryStatements));
+ return tx().query().insert(String.join("\n", insertQueryStatements)).collect(Collectors.toList());


It's more correct to try collecting it even if we don't check the value in the end, because we could skip some of the fails. However, it's more like a task for BDD to call for "check", so I'm not really sure if we need to collect on the "Given" step... Ready to roll back and make the same move for C# if you say so.

Given doesn't really mean anything in this case and may as well be Step. The actual step keywords are ignored.
.collect() could be replaced with .next() if it weren't for the fact that this "step" is reused as a helper function by other steps. As things stand though, this seems to be a reasonable compromise.

I'm okay with it as long as you are.

java/test/behaviour/query/language/rulevalidation/BUILD

java/test/behaviour/query/language/rulevalidation/RuleValidationTest.java

farost · 2024-03-15T16:07:39Z

rust/tests/behaviour/query/reasoner/steps.rs

@@ -103,6 +103,6 @@ generic_step_impl! {

 #[step(expr = "verify answers are consistent across {int} executions")]
 async fn verify_answers_are_consistent_across_executions(_context: &mut Context, _executions: usize) {
- // We can't execute previous query again because don't remember the query
+ // We can't execute previous query again because don't remember the query // TODO: start saving the query and implement


Maybe we want to create a task AND reference this task from the code so these TODOs could be found more easily?

Cool, thank you! What do you think about starting referencing github issues inside the code TODOs? For example:
// TODO #612: start saving the query and implement

It would help us cleaning outdated todos and remembering reasons behind each of them on a long distance.

csharp/Connection/TypeDBSession.cs

dmitrii-ubskii

Partial review

c/swig/typedb_driver_csharp.swg

dmitrii-ubskii · 2024-03-18T11:55:08Z

cpp/test/behaviour/steps/common/common_steps.cpp

@@ -32,7 +32,7 @@ cucumber_bdd::StepCollection<Context> commonSteps = {
 BDD_STEP("wait (\\d+) seconds", {
 std::this_thread::sleep_for(std::chrono::seconds(atoi(matches[1].str().c_str())));
 }),
- BDD_NOOP("set time-zone is: (.*)"),
+ BDD_NOOP("set time-zone is: (.*)"), // TODO: Decide if we want to implement (can be changed only for POSIX).


Probably okay for now.

dmitrii-ubskii · 2024-03-18T12:04:09Z

csharp/Api/BUILD

+
+
+csharp_library(
+ name = "api",


Should the target names follow the package names and also be named in UpperCamelCase?

It is actually a good point!

However, it looks like (if we want to follow it...) it's even worse: following the guideline, our targets (which are dlls) should be like Typedb.Driver.*** as well

Could be especially confusing in terms of tests: Typedb.Driver.Behaviour.Connection.Database.Steps, Typedb.Driver.Behaviour.Connection.Database.TestCore...

Not sure if we want, although it would be better to make user-facing libs following the convensions.

I think the 'right' convention should be to use our Bazel convention for targets, and set a separate name for any libraries we produce that are 'imported' afterward? ie. in java we can call the package API, but what matters is the package declaration at the top of the file... is that something that happens in C#?

It looks like I can't specify a separate name for the library file using the rule ([impl](https://github.com/bazelbuild/rules_dotnet/blob/862c7f5b8bc9d13de064ef0183e30da11034f3a4/dotnet/private/rules/csharp/library.bzl#L4)). Moreover, these files are generated without namespaces/packages in their names. so its just driver-csharp.dll, api.dll, common.dll for now. So I'm not sure what to follow and if we want to try to rename these files with an additional rule to have TypeDB.Driver.Apiand... Uhm... TypeDB.Driver?

dmitrii-ubskii · 2024-03-18T12:06:07Z

csharp/Api/BUILD

+
+csharp_library(
+ name = "api",
+ srcs = glob(["*.cs", "*/*.cs", "*/*/*.cs"]),


This can be **/*.cs, which is recursive.

Thanks! I copied what Java does, and this style is more explicit for the architecture's understanding, I guess. I'd leave it like this, but I'm ready to refactor it if you want me to.

Looks OK as is!

dmitrii-ubskii · 2024-03-18T12:09:11Z

csharp/Api/Concept/IConcept.cs

+ IType AsType()
+ {
+ throw new TypeDBDriverException(
+ ConceptError.INVALID_CONCEPT_CASTING, this.GetType().Name, typeof(IType).Name);


What does that mean for us?
Is typeof(this) going to be different from this.GetType(), i.e. IConcept rather than EntityTypeImpl?

dmitrii-ubskii · 2024-03-18T12:17:25Z

csharp/Api/Concept/Value/IValue.cs

+
+ public class ValueType : NativeObjectWrapper<Pinvoke.ValueType>
+ {
+ public static readonly ValueType OBJECT = new ValueType(typeof(object), false, false, Pinvoke.ValueType.Object);


Can we associate the variants with native ValueTypes here, and turn ValueClass etc. into get properties?

csharp/Api/Concept/Value/IValue.cs

dmitrii-ubskii · 2024-03-18T12:21:32Z

csharp/Api/Database/IDatabaseManager.cs

+ * driver.Databases.All;
+ * </pre>
+ */
+ IList<IDatabase> All { get; }


In our examples it's usually written as:

IDatabaseManager databases = driver.Databases; databases.All;

Which is a bit better.

dmitrii-ubskii · 2024-03-18T12:25:38Z

csharp/Api/Answer/IConceptMap.cs

+ * conceptMap.Variables;
+ * </pre>
+ */
+ IEnumerable<string> Variables { get; }


If it's a network call, it should probably be a method as well, to make it obvious that it's not free. Ex.: IDatabaseManager.All

csharp/Api/TypeDBCredential.cs

farost · 2024-03-18T13:39:56Z

java/test/behaviour/query/language/rulevalidation/RuleValidationTest.java

My previous comment got outdated for some reason, but I'll repeat it: this test was missing in Java!

dmitrii-ubskii · 2024-03-18T15:14:13Z

csharp/Api/Concept/Type/IType.cs

+ */
+ Promise<bool> IsDeleted(ITypeDBTransaction transaction);
+ }
+}


⛔️❗️

dmitrii-ubskii · 2024-03-18T15:14:31Z

csharp/Api/Concept/Type/IType.cs

+ */
+ Promise<bool> IsDeleted(ITypeDBTransaction transaction);
+ }
+}


Check IDEA settings.

dmitrii-ubskii · 2024-03-18T15:15:09Z

csharp/Api/Concept/Value/IValue.cs

+
+ public class ValueType : NativeObjectWrapper<Pinvoke.ValueType>
+ {
+ public static readonly ValueType OBJECT = new ValueType(typeof(object), false, false, Pinvoke.ValueType.Object);


Are Pinvoke.ValueTypes not integers?

dmitrii-ubskii · 2024-03-18T17:09:10Z

csharp/Test/BUILD

Can't help but notice there aren't any changes in .factory/automation.yml, that's something that should be sorted out soonish.

dmitrii-ubskii · 2024-03-18T17:25:25Z

rust/tests/behaviour/query/reasoner/steps.rs

@@ -103,6 +103,6 @@ generic_step_impl! {

 #[step(expr = "verify answers are consistent across {int} executions")]
 async fn verify_answers_are_consistent_across_executions(_context: &mut Context, _executions: usize) {
- // We can't execute previous query again because don't remember the query
+ // We can't execute previous query again because don't remember the query // TODO: start saving the query and implement


java/test/behaviour/query/language/rulevalidation/RuleValidationTest.java

dmitrii-ubskii · 2024-03-18T17:30:11Z

java/test/behaviour/query/QuerySteps.java

 TypeQLInsert typeQLQuery = TypeQL.parseQuery(String.join("\n", insertQueryStatements));
- return tx().query().insert(String.join("\n", insertQueryStatements));
+ return tx().query().insert(String.join("\n", insertQueryStatements)).collect(Collectors.toList());


Given doesn't really mean anything in this case and may as well be Step. The actual step keywords are ignored.
.collect() could be replaced with .next() if it weren't for the fact that this "step" is reused as a helper function by other steps. As things stand though, this seems to be a reasonable compromise.

dmitrii-ubskii · 2024-03-18T17:31:58Z

csharp/nuget/BUILD

Should this package be moved under "dependencies/nuget", next to "dependencies/maven"?

I dunno, maybe! I put it inside csharp as it's a csharp-only thing, but if you want to store such dependencies in a specific place, I'm not against moving it.

farost · 2024-03-20T13:02:31Z

csharp/Common/Exception/ErrorMessage.cs

+{
+ public abstract class ErrorMessage
+ {
+ private readonly string _codePrefix;


(duplicate old comment as it was outdated within the refactoring):

readonly vs { get; }:

*If I expect to always get the same result for each call, it's a readonly field.
*If I expect that the result can be different, it is a property with a getter (with no setters).
*If we call a native function under the hood, it's a property with a getter (with no setters).

flyingsilverfin

Very good already, almost there!
Note: your integration tests are very complete, hopefully they didn't take you too long and were useful during development.

WORKSPACE

flyingsilverfin · 2024-03-21T13:20:52Z

c/swig/typedb_driver_csharp.swg

+ public $csclassname Released()
+ {
+ var cPtr = swigCPtr.Handle;
+ if (swigCMemOwn)


Is it possible to have a race condition here? Can the memory be released by two threads concurrently?

Didn't think about it! I can wrap the body of this method in lock(this) {} just as the SWIG-generated Dispose does it just not to worry about it.

protected virtual void Dispose(bool disposing) { lock(this) { if (swigCPtr.Handle != global::System.IntPtr.Zero) { if (swigCMemOwn) { swigCMemOwn = false; typedb_driverPINVOKE.delete_Session(swigCPtr); } swigCPtr = new global::System.Runtime.InteropServices.HandleRef(null, global::System.IntPtr.Zero); } } }

@dmitrii-ubskii what are you thoughts?

lock(this) { ... } in Release() is sensible iff every place that directly accesses swigCMemOwn or swigCPtr (like getCPtr() or equivalent) also locks on this.

Well, not every place, just Dispose. swigRelease does not, that's why I am unsure.

In that case Release() twice without a lock doesn't introduce extra races, because those were already present for all native calls.
Illustration for why the lock wouldn't help here: https://dotnetfiddle.net/kO2zJf

Well, yeah, you're right about the races and locks, looks like it already works this way, I guess...

c/swig/typedb_driver_csharp.swg

flyingsilverfin · 2024-03-21T13:24:48Z

c/swig/typedb_driver_csharp.swg

+%promise(String, string, string)
+%promise(Bool, bool, bool)
+
+/* void promises require special handling */


Note: you can use code blocks (single or triple ticks) to write whatever you want:

TemplatedClass<void>

flyingsilverfin · 2024-03-21T13:25:54Z

c/swig/typedb_driver_csharp.swg

+%}
+
+%typemap(csdisposing, methodname="Dispose", methodmodifiers="protected", parameters="bool disposing") VoidPromise %{{
+ lock(this)


Is this to prevent race conditions? What's thie method Dispose versus Released at the top? They look quite similar!

Dispose is auto generated by SWIG and is a method of freeing unmanaged resources when leaving using blocks with allocated memory for IDisposable interface. You see it here as I use %typemap(csbody) typemap to rewrite the body of the generated class, thus erasing this automatically generated method, so I just return it here.

But yes, they want to prevent race conditions here (with a bit dirty way, tbh).

On the other hand, Released is a method for manual release of our resources in specific situations, similar to Java Driver.

csharp/BUILD

flyingsilverfin · 2024-03-21T14:05:47Z

csharp/nuget/update.sh

+# 3. Check the fetched packages in the regenerated 'paket.csharp_deps.bzl' file.
+# 4. Reference this packages in your BUILD files via '@paket.csharp_deps//{package.name.in.lowercase}'.
+
+paket install


So this means that CSharp build is not hermitic right? We should outline this in the PR description and note the build requirements on the local system.

Well, it depends on what we consider as hermetic. You're right that this file is not hermetic (I forgot about it), but it is needed only if you want to run paket2bazel (read as "run update.sh"), which is needed only for developers when you update dependencies. The driver and its build afterwards is hermetic and I've already "tested" it while setting up my Linux env: I just needed to install Bazel and fetch the repo.

I see - build is hermitic, but updating dependencies is not. Correct?

java/api/concept/value/Value.java

flyingsilverfin · 2024-03-21T14:08:13Z

csharp/Test/Integration/Memory/MemoryTest.cs

+ public void SendArraysToNativeSide()
+ {
+ string dbName = "mydb";
+ ITypeDBDriver driver = Utils.OpenConnection();


Note: in your PR description, you should also use driver as the name for a variable holding a driver instance.

I do, it's inside the using!

csharp/Test/Integration/Memory/MemoryTest.cs

dmitrii-ubskii · 2024-03-22T11:48:57Z

c/swig/typedb_driver_csharp.swg

+ public $csclassname Released()
+ {
+ var cPtr = swigCPtr.Handle;
+ if (swigCMemOwn)


lock(this) { ... } in Release() is sensible iff every place that directly accesses swigCMemOwn or swigCPtr (like getCPtr() or equivalent) also locks on this.

dmitrii-ubskii · 2024-03-22T11:52:24Z

.factory/automation.yml

- python3.9 -m pip install -U cffi
- export SYNC_DEPENDENCIES_TOKEN=$REPO_GITHUB_TOKEN
- bazel run @vaticle_dependencies//tool/sync:dependencies -- --source ${FACTORY_REPO}@${FACTORY_COMMIT}
+ sudo add-apt-repository ppa:deadsnakes/ppa


From #610

Suggested change

sudo add-apt-repository ppa:deadsnakes/ppa

sudo add-apt-repository -y ppa:deadsnakes/ppa

You're right, it has been lost within rebases. Sorry for that

.factory/automation.yml

dmitrii-ubskii · 2024-03-22T11:56:34Z

c/swig/typedb_driver_csharp.swg

+ arrayPtr[i] = global::System.Runtime.InteropServices.Marshal.StringToCoTaskMemAnsi($csinput[i]);
+ }
+
+ arrayPtr[i] = global::System.IntPtr.Zero;


This should really just be arrayPtr[arraySize] so it's both explicit and we don't have to depend on the for-loop side effects.

dmitrii-ubskii · 2024-03-22T12:16:53Z

cpp/lib/common/error_message.cpp

@@ -32,7 +32,7 @@ namespace TypeDB {

 namespace DriverError {

-#define ERR_DRIVER(ID, MSG) ERRMSG("CCL", "Driver Error", ID, MSG)
+#define ERR_DRIVER(ID, MSG) ERRMSG("CCCL", "Driver Error", ID, MSG)


@flyingsilverfin @farost The reason we introduced language prefixes in codes is so that we don't run into collisions between the Rust and the host language's error codes, mainly for internal errors. INT02 coming from Java's ""Illegal state has been reached!" is distinct from Rust INT02 "Unable to send response over callback channel (receiver dropped)."
We then reasoned that there's no guarantee that Rust driver won't be emitting its own Concept errors, e.g., so the prefix got added to all Java error codes. That should have also been the case in Python. Node is the only place where collisions are irrelevant as it's not using FFI.

farost · 2024-03-22T12:48:44Z

cpp/lib/common/error_message.cpp

@@ -32,7 +32,7 @@ namespace TypeDB {

 namespace DriverError {

-#define ERR_DRIVER(ID, MSG) ERRMSG("CCL", "Driver Error", ID, MSG)


Git went crazy, so I duplicate the conversation here.

The latest comment from @dmitrii-ubskii:

The reason we introduced language prefixes in codes is so that we don't run into collisions between the Rust and the host language's error codes, mainly for internal errors. INT02 coming from Java's ""Illegal state has been reached!" is distinct from Rust INT02 "Unable to send response over callback channel (receiver dropped)."
We then reasoned that there's no guarantee that Rust driver won't be emitting its own Concept errors, e.g., so the prefix got added to all Java error codes. That should have also been the case in Python. Node is the only place where collisions are irrelevant as it's not using FFI.

I had a talk about it with @flyingsilverfin. We thought that we are okay with the collisions of the error codes as far as they are just ideological, not real collision problems within the code (each driver runs separately). And for each error code we have context of the driver if somebody wants to check what to do in each specific situation.

I'm okay with two ways:

Having language-based prefix for all error codes we have in drivers

Not having one.

I don't really like exceptions like Nodejs, because the need of differentiating FFI/non-FFI drivers only complicates things here and a user receiving such errors is not interested if a driver they are using is based on FFI or not.

Are there technical problems with having same error codes with different content for different drivers?

We have checks in cloud tests that compare error codes.

UX and error reporting suffers when users search for the error code and find something entirely different.

We do not have checks that compare error codes, they would fail on factory otherwise. Only steps with "exception contains substring"

If we care about UX, then we still have to change python and nodejs codes as well (they were outdated).

@flyingsilverfin has @dmitrii-ubskii been convincing enough to accept change to "CCDR" (cpp), "CSDR" (csharp), "JDR" (java), "PDR" (python), "NDR" (node?..) (or "TDR" as typescript?..)?

I mean yeah i don't mind using a unique prefix for each! We can do that :)
C++ - CCDR
C# - CSDR
Java - JDR
Rust - RDR
Node - NDR
Python - PDR

If C has its own then we can use CDR, but i think it doesn't?

Yeah, it does not. Okay, I proceed with this plan then, thanks!

We do not have checks that compare error codes, they would fail on factory otherwise. Only steps with "exception contains substring"

Sorry, I meant in the typedb-cloud repo.

…#623) ## Usage and product changes We introduce packaging, distribution and documentation C# driver for TypeDB ([the original driver PR](#609)). It is built using the cross-platform [.NET 6 framework](https://dotnet.microsoft.com/en-us/download/dotnet/6.0). ### Usage: The driver is distributed as a series of [Nuget](https://www.nuget.org) packages. To use the driver, import the latest versions of the driver (TypeDB.Driver) and its Pinvoke runtime (TypeDB.Driver.Pinvoke) suitable for your platform. **CS project**: Here is an example from a `.csproj` for MacOS x86-64: ```xml <PackageReference Include="TypeDB.Driver" Version={VERSION} /> <PackageReference Include="TypeDB.Driver.Pinvoke.osx-x64" Version={VERSION} /> ``` If you aim to build a platform-independent package, reference all the needed runtimes (it will affect the size of your application by downloading a respective set of platform-specific dynamic libraries): ```xml <PackageReference Include="TypeDB.Driver.Pinvoke.osx-x64" Version={VERSION} /> <PackageReference Include="TypeDB.Driver.Pinvoke.linux-x64" Version={VERSION} /> <PackageReference Include="TypeDB.Driver.Pinvoke.win-x64" Version={VERSION} /> <PackageReference Include="TypeDB.Driver.Pinvoke.osx-arm64" Version={VERSION} /> ... ``` **Bazel**: 1. Import both the `TypeDB.Driver` and `TypeDB.Driver.Pinvoke` nuget packages as dependencies to your build rule. 2. For development, you can also use a [csharp bazel rule](https://github.com/bazelbuild/rules_dotnet/tree/master), passing targets `//csharp:driver-csharp` (the driver itself), `//csharp/Api:api` (exposed Api namespace), `//csharp/Common:common` (exposed Common namespace) as dependencies. **A simple usage example** (see `csharp/Test/Integration/Examples` for more): ``` using TypeDB.Driver.Api; using TypeDB.Driver.Common; public class TypeDBExample { public void SetupTypeDB() { string dbName = "access-management-db"; string serverAddr = "127.0.0.1:1729"; try { using (ITypeDBDriver driver = Drivers.CoreDriver(serverAddr)) { driver.Databases.Create(dbName); IDatabase database = driver.Databases.Get(dbName); using (ITypeDBSession session = driver.Session(dbName, SessionType.Schema)) { using (ITypeDBTransaction transaction = session.Transaction(TransactionType.Write)) { transaction.Query.Define("define person sub entity;").Resolve(); string longQuery = "define name sub attribute, value string; person owns name;"; transaction.Query.Define(longQuery).Resolve(); transaction.Commit(); } } database.Delete(); } } catch (TypeDBDriverException e) { Console.WriteLine($"Caught TypeDB Driver Exception: {e}"); } } } ``` ## Implementation **Documentation:** Documentation for C# is generated in a similar way as the C/C++ ones: `doxygen` -> `adoc`. This way, the C# parser has some extra logic of parsing the intermediate doxygen files but follows a similar (copypasted) structure. Additionally, some small bugs for both old parsers have been fixed. **Distribution:** We decided to distribute the C# driver as a series of packages to allow the end user to choose between multi-platform and compactness. This way, we have: `TypeDB.Driver.{X-X-X-version}.nupkg` - the main package with the driver, which fails in runtime if no runtime is additionally provided. `TypeDB.Driver.Pinvoke.{X-X-X-version}.{platform}.nupkg` - platform-specific runtimes (`osx-x64`, `osx-arm64`, `linux-x64`, `linux-arm64`, `win-x64`). The distribution is implemented via several bazel rules: `nuget_pack` (with a wrapper for native packages in this PR) and `nuget_push` and their respective calls for both runtime and main packages. **Tests:** Integration tests use NUnit framework as a built-in framework for C# bazel test rules. This PR presents both `test-core` and `test-cloud` as small usage examples within the integration tests. These tests are also triggered by Factory. There is also a new Deployment test, which creates a `.csproj` project and tries to fetch all the needed (and all the available runtime) packages to use them for another small C# program. This test is triggered by CircleCI.

farost added the type: feature label Mar 15, 2024

farost requested a review from alexjpwalker March 15, 2024 12:16

farost requested review from dmitrii-ubskii and flyingsilverfin as code owners March 15, 2024 12:16

typedb-bot assigned alexjpwalker Mar 15, 2024

typedb-bot assigned dmitrii-ubskii and flyingsilverfin Mar 15, 2024

farost mentioned this pull request Mar 15, 2024

Dependencies update for C# driver. typedb/dependencies#515

Merged

farost commented Mar 15, 2024

View reviewed changes

csharp/Connection/TypeDBSession.cs Outdated Show resolved Hide resolved

dmitrii-ubskii requested changes Mar 18, 2024

View reviewed changes

farost commented Mar 18, 2024

View reviewed changes

dmitrii-ubskii requested changes Mar 18, 2024

View reviewed changes

farost commented Mar 20, 2024

View reviewed changes

flyingsilverfin requested changes Mar 21, 2024

View reviewed changes

flyingsilverfin approved these changes Mar 21, 2024

View reviewed changes

farost mentioned this pull request Mar 21, 2024

Ci sync deps #610

Closed

farost requested a review from dmitrii-ubskii March 21, 2024 16:37

farost force-pushed the csharp-kickoff branch 7 times, most recently from 95d1436 to 914ff51 Compare March 21, 2024 17:55

farost added 2 commits March 21, 2024 18:30

Add hello world with working Bazel-Dotnet setup

2bee364

Divide lib and binary for the Hello World version

4bd4cfe

farost added 17 commits March 21, 2024 18:43

Exclude python from build (temporarily)

b4ddb7b

Try to fix python ignoring

555ce59

Fix dependencies

fab4cba

Disable cpp for space

0042590

Make enum-like classes actual enums (ValueType WIP)

fbe134b

Fix enums and return main build

410c819

Fix checkstyle for nuget package

48dde68

Add docs for ValueType methods. Fix nuget package build

1151b08

Fix checkstyle comments

c95f21f

Squeeze .factory steps declarations. Fix TestRunner float fails

7d9f68d

Fix docs for java and python

933531b

Refactor Properties <-> Methods. Fix some documentation

0be1953

Fix no empty line in the end

ac7577f

Change Concepts -> GetConcepts

513a47e

Fix linux memory marshalling

2af3628

Refactor error messages codes. Update deps

08a6275

Fix automation conflict bugs

0af6bb7

farost force-pushed the csharp-kickoff branch from 4b663db to 0af6bb7 Compare March 21, 2024 18:43

Fix .factory

c9aeaa9

dmitrii-ubskii reviewed Mar 22, 2024

View reviewed changes

farost commented Mar 22, 2024

View reviewed changes

Refactor error codes. Refactor memory management

02f2d0e

farost requested a review from dmitrii-ubskii March 22, 2024 13:54

dmitrii-ubskii approved these changes Mar 22, 2024

View reviewed changes

farost merged commit a3041f1 into typedb:development Mar 22, 2024
23 checks passed

farost deleted the csharp-kickoff branch March 22, 2024 15:20

farost restored the csharp-kickoff branch March 25, 2024 08:58

farost mentioned this pull request Mar 25, 2024

Introduce C# driver without documentation and deployment #616

Merged

farost mentioned this pull request Apr 4, 2024

Introduce packaging, distribution and documentation for the C# driver #623

Merged

	sudo add-apt-repository ppa:deadsnakes/ppa
	sudo add-apt-repository -y ppa:deadsnakes/ppa

		@@ -32,7 +32,7 @@ namespace TypeDB {

		namespace DriverError {

		#define ERR_DRIVER(ID, MSG) ERRMSG("CCL", "Driver Error", ID, MSG)

Introduce C# driver without deployment #609

Introduce C# driver without deployment #609

Conversation

farost commented Mar 15, 2024 • edited Loading

Usage and product changes

Implementation

typedb-bot commented Mar 15, 2024 • edited by flyingsilverfin Loading

PR Review Checklist

farost commented Mar 15, 2024 • edited Loading

A quick comment about Naming

farost left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

dmitrii-ubskii left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

flyingsilverfin left a comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

farost Mar 21, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

farost Mar 21, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

farost Mar 22, 2024 • edited Loading

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

Choose a reason for hiding this comment

farost commented Mar 15, 2024 •

edited

Loading

typedb-bot commented Mar 15, 2024 •

edited by flyingsilverfin

Loading

farost commented Mar 15, 2024 •

edited

Loading

farost Mar 21, 2024 •

edited

Loading

farost Mar 21, 2024 •

edited

Loading

farost Mar 22, 2024 •

edited

Loading