Using Code Coverage in Business Central Development

Intro

Sample code coverage summary

In the latest version of AL Test Runner I’ve added an overall percentage code coverage and totals for number of lines hit and number of lines. I’ve hesitated whether to add this in previous versions. Let me explain why.

Measuring Code Coverage

First, what these stats actually are. From right to left:

Code Coverage 79% (501/636)
  1. The total number of code lines in objects which were hit by the tests
  2. The total number of lines hit by the tests
  3. The percentage of the code lines hit in objects which were hit at least once

Notice that the stats only include objects which were hit by your tests. You might have a codeunit with thousands of lines of code, but if it isn’t hit at all by your tests it won’t count in the figures. That’s just how the code coverage stats come back from Business Central. Take a look at the file that is downloaded from the test runner if you’re interested (by default it’s saved as codecoverage.json in the .altestrunner folder).

It is important to bear this is mind when you are looking at the headline code coverage figure. If you have hundreds of objects and your tests only hit the code in one of them, but all of the code in that object – the code coverage % will be a misleading 100%. (If you don’t like that you’ll have to take it up with Microsoft, not me).

What Code Coverage Isn’t Good For

OK, but assuming that my tests hit at least some of the code in the most important objects then the overall percentage should be more or less accurate right? In which case we should be able to get an idea of how good the tests are for this project? No.

Code Coverage ≠ Test Quality

The fact that one or more tests hits a block of code does not tell you anything about how good those tests are. The tests could be completely meaningless and the code coverage % alone would not tell you. For example;

procedure CalcStandardDeviation(Values: List of [Decimal]): Decimal
var
    Value, Sum, Mean, SumOfVariance : Decimal;
begin
    foreach Value in Values do
        Sum += Value;
    Mean := Sum / Values.Count();
    foreach Value in Values do
        SumOfVariance += Power((Value - Mean), 2);
    exit(SumOfVariance / Values.Count());
end;

[Test]
procedure TestCalcStandardDeviation()
var
    Values: List of [Decimal];
begin
    Values.Add(1);
    Values.Add(3);
    Values.Add(8);
    Values.Add(12);

    CalcStandardDeviation(Values);
end;

Code coverage? 100% ✅

Does the code work? No ❌ The calculation of the standard deviation is wrong. It is a pointless test, it executes the code but doesn’t verify the result and so doesn’t identify the problem. (In case you’re wondering the result should be the square root of SumOfVariance).

Setting a Target for Code Coverage

What target should we set for code coverage in our projects? Don’t.

Why not? There are a couple of good reasons.

  1. There is likely to be some code in your project that you don’t want to test
  2. You might inadvertently encourage some undesired behaviour from your developers

Why Wouldn’t You Test Some of Your Code?

Personally, I try to avoid testing any code on pages. Tests which rely on test page objects take significantly longer to run, they can’t be debugged with AL Test Runner and I try to minimise the code that I write in pages anyway. Usually I don’t test any of:

  • Code in action triggers
  • Lookup, Drilldown, AssistEdit or page field validation triggers
  • OnOpen, OnClose, OnAfterGetRecord
  • …you get the idea, any of the code on a page

You might also choose not to test code that calls a 3rd party service. You don’t want your tests to become dependent on some other service being available, it is likely to slow the test down and you might end up paying for consumption of the service.

I would test the code that handles the response from the 3rd party but not the code that actually calls it e.g. not the code that sends the HTTP request or writes to a file.

Triggers in Install or Upgrade codeunits will not be tested. You can test the code that is called from those triggers, but not the triggers themselves.

Developing to a Target

When a measure becomes a target, it ceases to be a good measure.

Marilyn Strathern

If we already know that we have some code that we will not write tests for then it doesn’t make a lot of sense to set a hard target of 100%. But, what other number can you pick? Imagine two apps:

  1. An app that is purely responsible for handling communication with some Azure Functions. Perhaps the majority of the code in that app is working with HTTP clients, headers and responses. It might not be practical to achieve code coverage of more than 50%
  2. An app that implements a new sales price mechanism. It is pure AL code and the code is almost entirely in codeunits. It might be perfectly reasonable to expect code coverage of 95%

It doesn’t make sense to have a headline target for the developers to work to on both projects. Let’s say we’ve agreed as a team that we must have code coverage of at least 75%. We might incentivise developers on the first project to write some nonsense tests just to artificially boost the code coverage.

Meanwhile on the second project some developers might feel safe skipping writing tests for some important new code because the code coverage is already at 80%.

Neither of these scenarios is great, but, in fairness, the developers are doing what we’ve asked them to.

What is Code Coverage Good For?

So what is code coverage good for? It helps to identify objects that have a lot of lines which aren’t hit by tests. That’s why the output is split by object and includes the path to the source file. You can jump to the source file with Alt+Click.

Highlight the lines which were hit by the previous test run with the Toggle Code Coverage command. That way you can make an informed opinion about whether you ought to write some more tests for this part of the code or whether it is fine as it is.

50% code coverage might be fine when 1 out of 2 lines has been hit. It might not be fine when 360 out of 720 lines have been hit – but that’s for you to decide.

Further Reading

https://martinfowler.com/bliki/TestCoverage.html

Show Table Data from AL Test Runner

Why?

If you are writing Business Central automated tests then you are probably creating some data as part of those tests. Sometimes it can be useful to see the records that have been created, especially if you are trying to create the correct scenario for the GIVEN part of your test and are experimenting with the library codeunits in that part of the system.

Of course you can debug your tests and most of the time being able to see the variable values in the debugger is all you need. Other times it would be better to be able to see the whole table e.g. “why were there no Reservation Entries in the filter?” Opening the table, applying the filters and taking a look yourself can be faster than trawling through the debugger and tweaking the test code. See here for a little background.

What?

While you are debugging your test you can open a new web client session and navigate directly to the table by adding &table=[table id] to the end of the URL. You don’t need anything just to do that – just a browser. v0.5.4 of AL Test Runner adds a new command to launch the browser to the correct table id from the code. Just right click on the variable name and choose Show Table Data.

Table Locks

Depending on what data you are reading and when you are reading it you might find that the table is locked and the web client will not open it. There are a few things you can do in that case:

  • Try stepping through the code until the lock has been released on the table that you are interested in
  • Add a Commit after the data has been inserted (don’t worry – even code that is explicitly committed during a test is rolled back at the end)
  • Use a SQL query to read uncommitted data

Update Test Runner Service

You need v0.1.2 of the Test Runner Service app (because it adds a new method to return the table id for a given table name). You can install it using the Install Test Runner Service command in VS Code. If you’ve already got an older version installed you’ll need to start the app data upgrade yourself to complete the installation. You can get the app from here: https://github.com/jimmymcp/test-runner-service/raw/master/James%20Pearson_Test%20Runner%20Service.app

Video

If you prefer your demos prefaced with questionable accents I’ve got you covered here…

Debugging Business Central Tests with AL Test Runner

TL;DR

  1. Install the Test Runner Service app (see https://github.com/jimmymcp/test-runner-service; direct download of the app file from here) or use the “Install Test Runner Service” command from VS Code to install into the Docker container specified in the config file
  2. Set the URL to the test runner service in the testRunnerServiceUrl key of the AL Test Runner config file
  3. Define a debug configuration of request type ‘attach’ in launch.json to attach the debugger to the service tier that you want to debug (should be the same service tier as specified by the testRunnerServiceUrl key)

Overview

From v0.4.0 of the AL Test Runner app it is possible to debug Business Central tests without leaving Visual Studio Code. There’s a lot of scope for improvements but if you’re interested in trying it out it’s included in the marketplace version now.

Test Runner Service App

This is a very simple app that exposes a codeunit as a web service to accept a codeunit ID and test name to run. Those values are passed to a test runner codeunit (codeunit isolation) to actually run the tests. This is so that the tests are executed in a session type of WebService which the debugger can attach to (the PowerShell runner creates a session type of ClientService).

The app is in the per tenant object range: 79150-79160 to be precise (a number picked pretty much at random). If that clashes with some other object ranges present in the database you can clone the repo and renumber the codeunits if you want. The source is here: https://github.com/jimmymcp/test-runner-service

You can use the Install Test Runner Service command in VS Code to automatically download the app and install into the container specified in the AL Test Runner config file.

The app is not code signed so you’ll need to use the -SkipVerification switch when you install it.

testRunnerServiceUrl

A new key is required in the AL Test Runner config file. This specifies the OData endpoint of the test runner service that is exposed by the Test Runner Service app. The service will be called from the VS Code terminal – so consider where the terminal is runner and where the service is hosted.

We develop against local Docker containers so the local VS Code instance will be able to access the web service without any trouble. If you develop against a remote Docker host make sure that the OData port is available externally. If you use VS Code remote development remember that the PowerShell session will be running on the VS Code server host.

The url will be in the format:

http[s]://[BC host]:[OData port]/[BC instance]/ODataV4/TestRunner_RunTest?company=[BC company]

for example against a local Docker container called bc with OData exposed on the default port of 7048 and a company name of CRONUS International Ltd.:

"testRunnerServiceUrl": "http://bc:7048/BC/ODataV4/TestRunner_RunTest?company=CRONUS%20International%20Ltd."

Debug Configuration

You will need a debug configuration of type attach in the launch.json file. This should attach the debugger to the same service as identified by the testRunnerServiceUrl key. breakOnNext should be set WebServiceClient. Currently UserPassword authentication is the only authentication method supported.

{
    "name": "Attach bc",
    "type": "al",
    "request": "attach",
    "server": "http://bc",
    "serverInstance": "bc",
    "authentication": "UserPassword",
    "breakOnError": true,
    "breakOnRecordWrite": false,
    "enableSqlInformationDebugger": true,
    "enableLongRunningSqlStatements": true,
    "longRunningSqlStatementsThreshold": 500,
    "numberOfSqlStatements": 10,
    "breakOnNext": "WebServiceClient"
}

Debugging

Codelens actions will be added at the top of test codeunits and before each test method. Set a breakpoint in the test method that you want to debug or allow the debugger to break on an error.

Clicking on Debug Test (Ctrl+Alt+D) will attach the first debug configuration specified in launch.json and call the web service to run the test with the Test Runner Service app.

Attaching the debugger and running a test from VS Code

Step in/out/over as usual. When the code execution has finished if an error was encountered the error message and callstack will be displayed in the terminal.

Limitations

There are some limitations to running tests in a web service session. Most importantly TestPage variables are not supported. There may also be some differences in the behaviour of tests in web services and the PowerShell runner.

Testing Internal Functionality

Internal Access Modifier

We’ve had access modifiers in Business Central for a little while now. You can use them to protect tables, fields, codeunits and queries that shouldn’t be accessible to code outside your app.

For example, you might have a table that contains some sensitive data. Perhaps some part of a licensing mechanism or internal workings of your app that no one else should have access to. Mark the table as:

Access = Internal;

and only code in your app will be able to access it. Even if someone develops an app that depends on your app they will receive a compile error if they create a variable to the table: “<table> is inaccessible due to its protection level.” Before you ask about RecordRefs – I don’t know, I haven’t tested. I assume that Microsoft have thought of that and prevent another app from opening a RecordRef to an internal table belonging to another app.

Alternatively you might have a function in a codeunit that shouldn’t be called from outside your app. The function needs to be public so that other objects in your app can call it, but you can mark it as internal to prevent anyone else calling it:

internal procedure SomeSensitiveMethod()
begin
  //some sensitive code that shouldn't be accessible from outside this app
end;

Testing

Cool.

But wait…how do we test this functionality? We develop our tests alongside the app code but split the test codeunits out into a separate app in our build pipeline – because that’s how Microsoft like it for AppSource submissions.

The result is that the tests run fine against the local Docker container that I am developing and testing against. I push my changes to Azure DevOps to create a pull request and…the build fails. My (separate) test app is now trying to access the internal objects of the production app and fails to compile.

The solution is to use the internalsVisibleTo key in app.json of the production app. List one or more apps (by id, name and publisher) that are allowed to access the internals of the production app. More about that here.

Maybe you already develop your tests as a separate app and so can copy the app id from app.json of the test app.

In our case we usually generate a new guid for the test app as part of the build process – because we don’t usually care what id it has. For times we do want to specify the id of the test app we have an environment.json file that holds some settings for the build – Docker image, credentials, translations to test etc. We can set a testappid in that file and include it in the internalsVisibleTo key in app.json.

Now the build splits the apps into two and creates a test app with the id specified by testappid which compiles and can access internal objects and functions of the production app.

Performance of Test Code

Let’s talk about the performance of the test code that we write for Business Central. What do I mean by “performance” and how can we improve it?

Defining “Performance”

Obviously, before we set out to improve something we need to have an idea of what it is we’re trying to optimise for. I’m coming to think of the performance of test code in a couple of key ways:

  1. How easy/quick is it to write test code?
  2. How quickly do the tests run?

Performance of Writing Tests

I suppose none of the below points are specific to test code. They are relevant to any sort of code that we are writing but we can be more inclined to neglect them for test code than production code. If you embrace any sort of automated tested discipline you’re going to spend a significant proportion of your time reading and writing test code – perhaps even as much as you do on production code. It is well worth investing a little time to clean up the code and making it easier to read and maintain.

Comments

Say what you like about comments in production code – variable names should declare the intent, comments are evil blah blah – I do find a few comments valuable in a test.

In fact, I write them first. Given some set of circumstances, when this happens then this is the expected result. Writing those sentences first helps to be clear about what I’m trying to test and what the desired behaviour actually is.

Of course the code in between those comments should be readable and easy to follow – but if you are diligent with a few comments per test you can describe the expected behaviour of that part of the system without having to read any code.

Grouping Tests Logically

I appreciate the situation is different for VARs but as an ISV we have large object ranges to do our development in. There is no reason for us to bundle unrelated code into the same codeunit. If we are starting work on separate from existing business logic then it belongs in its own codeunit. In which case, why wouldn’t the corresponding tests also go in their own codeunit?

Ideally I want to be able to glance down the list of test codeunits that we have and see logical grouping of tests that correspond to recognisable entities or concepts in the production code. If I want to see how our app changes the behaviour of warehouse receipts I can look in WhseReceiptTests.Codeunit.al.

Refactoring Into Library Codeunits

As soon as you start writing tests you’ll start working with the suite of library codeunits provided by Microsoft. You’ll notice that they are separated into different areas of the system e.g. Library – Sales, Library – Warehouse, Library – Manufacturing and so on.

Very likely you’ll want to create your own library codeunit to:

  • Initialise some setup, perhaps create your app’s setup record, create some No. Series etc.
  • Create new records required by your tests – it is useful to follow the convention LibraryCodeunit.CreateXYZ(var XYZ: Record XYZ);
  • Consider having those Create functions returning the primary key of the new record as well so the result can be used inline in the test e.g.
    • LibraryCodeunit.CreateXYZ(var XYZ: Record XYZ): Code[20]
    • LibraryCodeunit.CreateXYZNo(): Code[20]
  • Use method overloads for the create functions – have an overload with more parameters when you need to specify extra field values but keep a simple overload for when you don’t
  • Identify blocks of code that are often required in tests and consider moving them to a library method e.g. creating an item with some bespoke fields populated in a certain way

Having a comprehensive library codeunit brings two benefits:

  1. Tests are easier and faster to write if you already have library methods to implement the Given part of the test
  2. Less code in the tests, making them easier to read

Performance of Running Tests

First, why do we care about how long tests take to run? Does it really matter if your test suite takes an extra minute or two to run?

Obviously, we want our builds to complete as quickly as possible, while still performing all the checks and steps that we want to include. The longer a build takes the more likely another one is going to be queued at the same time and eventually someone is going to end up having to wait. We’ve got a finite number of agents to run simultaneous builds (we host our own – more on that here if you’re curious).

But that isn’t the biggest incentive.

I’m a big fan of running tests while I’m developing – both new tests that I’m writing to cover my new code and existing tests (more on that here). I usually run:

  • the test that I’m working on very frequently (at least every few minutes – see it fail, see it pass)
  • all the tests in that codeunit frequently (maybe each time I’ve finished with a new test)
  • the whole test suite every so often (at least 2-3 times before pushing my work and creating a pull request)

After all, if you’ve got the tests, why not run them? I should know sooner rather than later if some code that I’ve changed has broken something. If the whole test suite takes 60 seconds to run that’s fine. If it takes 10 minutes that’s more of a problem.

In that case I’ll be more inclined to push my changes to the server without waiting, keep a build agent busy for half an hour, start working on something else and then get an email saying the build has failed. Something I could have realised and fixed if I’d run the test suite before I pushed my changes.

So, how to make them faster?

Minimal Setup

First, only create the scenario that is sufficient for your test. For example, we work with warehousing functionality a lot. If we’re testing something to do with warehouse shipments do we need a location with advanced warehousing, zones, bin types, bins, warehouse employees…?

Probably not. Likely I can create a location without Bin Mandatory or Requires Pick and still create a sufficient test.

If you need ledger entries to test with you may be able to create and post the relevant journals rather than creating documents and posting them. Creating an item ledger entry by posting an item journal line is faster than posting a sales order.

Or, you probably want to prevent negative inventory in real life – but does that matter for your test? Save yourself the trouble of having to post some inventory before shipping an item and just allow it to go negative.

Try to restrict the setup of your tests to what is actually essential for the scenario that you are testing. Answering that question is, in itself, a useful thought process.

Shared Fixture

Better yet, set something up once per test codeunit and reuse it in each of the tests. This what Luc van Vugt refers to as a “shared fixture”. You should check out his blog for more about that.

I feel a little mixed about this. I like the idea that each test is entirely responsible for creating its own given scenario and isn’t dependent on anything else but this is denying that this is faster. Finding a posted sales invoice that already exists is much faster than creating a customer, item, sales order and shipping and invoicing it.

No Setup

What is even faster than setting up some data one time? Doing it no times. Depending on what you are testing you may just be able to insert the records you need or call field validation on a record without inserting it.

If I’m testing that validating a field behaves in a certain way I may not need a record to actually be inserted in the table.

Alternatively if you need a sales invoice header to test with you might be able just to initialise it and call SalesInvHeader.Insert; It feels so wrong – but if your test just needs a record and doesn’t care where it came from, who cares? It will all be rolled back after the tests have run anyway.