21221
Comment:
|
22677
|
Deletions are marked like this. | Additions are marked like this. |
Line 9: | Line 9: |
The purpose of this class is to provide a framework for testing all the projects of the APLTree project. Only with such a framework is it possible to make changes to the test suite without too much hassle. | The purpose of this class is to provide a framework for testing all the projects of the APLTree library. Only with such a framework is it possible to make changes to any APLTree project with confidence. |
Line 21: | Line 21: |
1. The `#.Tester` class assumes that all your tests are hosted by an ordinary namespace. All methods running test cases take a ref to that namespace as an argument. 1. All test functions inside that namespace are expected to start their names with the string `Test_` followed by digits. '''Update''': since version 1.1 test functions can be grouped. For example, in order to group all functions testing a method `foo` the test functions are named `Test_foo_001` (or `Test_foo_01` or `Test_foo_00001`), `Test_foo_002`, `Test_foo_003` and so on. 1. The first line after the header must carry a comment telling what is actually tested. Keep in mind that later you might be in need for finding out which test cases are testing a certain method! |
1. The `#.Tester` class assumes that all your tests are hosted by a namespace. It must not be an unnamed namespace. All methods of the `Tester` class running test cases take a ref to that namespace as an argument. 1. All test functions inside that namespace are expected to start their names with the string `Test_` followed by digits. '''Update''': since version 1.1 test functions can be grouped. For example, in order to group all functions testing a method `foo` the test functions are named `Test_foo_001`, `Test_foo_002`, `Test_foo_003` and so on. 1. The number of digits you use for numbering is not restricted: `Test_foo_1` is as fine as is `Test_foo_0000001`. However, they should be consistent. When starting to implement test cases you are advised to leave gaps: `Test_foo_001`, `Test_foo_010` etc. 1. The first line after the header must carry a comment telling what is actually tested. Keep in mind that later this is the only way to tell one test case from the others without reading the code, so be as clear as you can possibly be. |
Line 25: | Line 26: |
* Run with error trapping and return a log (vector of text vectors) reporting details. Run this before checking your code into a Code Management System like !SubVersion. See `Tester.Run` for details. | * Run with error trapping and return a log (vector of text vectors) reporting details. Run this before checking your code into a Code Management System like !SubVersion, !GitHub or [[acre]]. See `Tester.Run` for details. |
Line 27: | Line 28: |
* Run "batch mode". It is assumed that those test cases that need a human being in front of the monitor are '''not''' executed when running in batch mode. See `Tester.RunBatchTests` for details. (Of course it is best not to rely on a human being but this cannot always be avoided) | * Run "batch mode". It is assumed that those test cases that need a human being in front of the monitor are '''not''' executed when running in batch mode. See `Tester.RunBatchTests` for details. (Of course it is best not to rely on a human being but this cannot always be avoided). There is also a method `RunBatchTestsInDebugMode` available. |
Line 31: | Line 32: |
1. Every test function must return a result. Use one of the following niladic functions as explicit result: | 1. Every test function must return a result. Use one of the following niladic functions (acting like constants) as explicit result: |
Line 53: | Line 54: |
First of all the `Run*` method checks whether there is a file `testcases_{computername}.ini`. If this is the case that INI file is processed. Use this to specify important computer-dependent variables. | First of all the `Run*` method checks whether there is a file `testcases_{computername}.ini`. If this is the case that INI file is processed. Use this to specify computer-dependent variables. |
Line 57: | Line 58: |
Note that if one of the two INI files exists there will be a flat namespace `{yourTestNamespace}.INI`, meaning that sections in the INI file are ignored. An example: if your test functions are hosted by a namespace "foo" and your INI file specifies a variable hello as holding "world" then: {{{ 'world'≡#.foo.INI.hello |
Note that if one of the two INI files exists there will be a flat namespace `{yourTestNamespace}.INI`, meaning that sections in the INI file are ignored. An example: if your test functions are hosted by a namespace `Foo` and your INI file specifies a variable `hello` as holding "world" then: {{{ 'world'≡#.Foo.INI.hello |
Line 68: | Line 69: |
Note that the function must be either niladic or monadic and must not return a result. If the function is monadic then a Boolean is passed via the right argument telling the function that the test suite is going to run in batch mode if true. | Note that the function must be either niladic or monadic and --(must not return a result)-- and, since version 3.0, may or may not return a Boolean result. A 1 means that function did what it is supposed to do (=same as no result) while a 0 means it could not initialize. Of course you can simply execute `→` on a single line in your `Initial` function if any requirement is not met but that would also mean that your test did not run at all. If you run your test case automatically somehow than it should return a 1 indicating failure. If the function is monadic then a Boolean is passed as the right argument telling the `Initial` function whether the test suite is going to run in batch mode (1) or not. ############ |
Line 104: | Line 111: |
PassesIf Usage : →PassesIf x, where x is a boolean scalar | PassesIf Usage : →PassesIf x, where x is a boolean scalar GetTestFnsTemplate ⍝ Returns a vector of text vectors with the code of a template test case. |
Line 108: | Line 116: |
⍝ Niladic functions returning an integer constant, see table above. | ∆... ⍝ Niladic functions returning an integer constant, see table above. |
Line 129: | Line 137: |
That's why the template for the test function carries such a statement '''and''' keeps `⎕TRAP` local'. | That's why the template for the test function carries such a statement '''and''' keeps `⎕TRAP` local. |
Line 134: | Line 142: |
→GoToTidyUp a≡b | →GoToTidyUp expected≢result |
Line 141: | Line 149: |
In case the namespace has no test functions in it (yet) the `EstablishHelpersIn` function also creates a test function template named `Test_000`. | In case the namespace has no test functions in it (yet) the `EstablishHelpersIn` function also creates a test function template named `Test_000`. ('''Not''' when the hosting namespace is scripted!) |
Line 157: | Line 165: |
The most important helper function is probably `L`: with a large number of test functions it's the only reasonable way to gather information. Since version 1.8.0 `L` requires a right argument (empty or group name) and an optional left argument (test case numbers). | The most important helper functions are probably `G` and `L`: with a large number of test functions it's the only reasonable way to gather information: `G` lists all groups and `L` all test functions. Since version 1.8.0 `L` requires a right argument (empty or group name) and an optional left argument (test case numbers). |
Line 172: | Line 182: |
L 'List' | L 'Li' |
Line 185: | Line 195: |
Note that you don't need to specify more characters than needed to actually identify a group. |
|
Line 205: | Line 217: |
* Try to keep your test cases simple and test just one thing at a time, for example just one method at a time, and only if that method is simple. | * Try to keep your test cases simple and test just one thing at a time, for example just one method at a time. If a method is complex you will be better off creating a group (use the method name as group name) with several simple test cases rather than one complex test case. |
Line 209: | Line 221: |
* Notice that the DRY principle (don't repeat yourself) can and should be ignored in test cases. == Methods == |
* Notice that the DRY principle (don't repeat yourself) can and should be ignored in test cases: any test case should read from top to bottom like an independent story. == Tester's methods == |
Tester
Contents
Tester is part of the CategoryAplTree project.
Overview
The purpose of this class is to provide a framework for testing all the projects of the APLTree library. Only with such a framework is it possible to make changes to any APLTree project with confidence.
You might find the framework flexible enough to suit your own needs when it comes to implementing tests, even independently from the APLTree project.
Details
Terminology
Note that test cases causing a crash are considered "broken". Test cases that do not return the expected result are considered "failing".
Assumptions and preconditions
The #.Tester class assumes that all your tests are hosted by a namespace. It must not be an unnamed namespace. All methods of the Tester class running test cases take a ref to that namespace as an argument.
All test functions inside that namespace are expected to start their names with the string Test_ followed by digits. Update: since version 1.1 test functions can be grouped. For example, in order to group all functions testing a method foo the test functions are named Test_foo_001, Test_foo_002, Test_foo_003 and so on.
The number of digits you use for numbering is not restricted: Test_foo_1 is as fine as is Test_foo_0000001. However, they should be consistent. When starting to implement test cases you are advised to leave gaps: Test_foo_001, Test_foo_010 etc.
- The first line after the header must carry a comment telling what is actually tested. Keep in mind that later this is the only way to tell one test case from the others without reading the code, so be as clear as you can possibly be.
- It is assumed that you have three different scenarios when you want to run test cases with a different behaviour:
Run with error trapping and return a log (vector of text vectors) reporting details. Run this before checking your code into a Code Management System like SubVersion, GitHub or acre. See Tester.Run for details.
Run without error trapping. Use this for investigating why test cases crash or fail. See Tester.RunDebug for details.
Run "batch mode". It is assumed that those test cases that need a human being in front of the monitor are not executed when running in batch mode. See Tester.RunBatchTests for details. (Of course it is best not to rely on a human being but this cannot always be avoided). There is also a method RunBatchTestsInDebugMode available.
- Every test function must accept a right argument which is a two-item vector of Booleans:
debugFlag (0=use error trapping)
batchFlag (0=does not run in batch mode)
- Every test function must return a result. Use one of the following niladic functions (acting like constants) as explicit result:
∆OK |
Passed |
∆Failed |
Unexpected result |
∆NoBatchTest |
Not executed because batchFlag was 1. |
∆InActive |
Not executed because the test case is inactive (not ready, buggy, whatever) |
∆LinuxOnly |
Not executed because runs under Linux only. |
∆LinuxOrMacOnly |
Not executed because runs under Linux/Mac OS only. |
∆LinuxOrWindowsOnly |
Not executed because runs under Linux/Windows only. |
∆MacOrWindowsOnly |
Not executed because runs under Mac OS/Windows only. |
∆MacOnly |
Not executed because runs under Mac OS only. |
∆NoAcreTests |
Not executed because it's acre related. |
∆WindowsOnly |
Not executed because runs under Windows only. |
These niladic functions are made available as Helpers - see there.
Work flow
No matter which of the Run* functions (Run, RunBatchTests, RunDebug, RunThese) you are going to call, the workflow is always the same:
INI files
First of all the Run* method checks whether there is a file testcases_{computername}.ini. If this is the case that INI file is processed. Use this to specify computer-dependent variables.
If no computer-specific INI file was found then it checks whether there is a file testcases.ini. If this is the case that INI file is processed, too. Use this to specify general stuff that does not depend on a certain computer/environment.
Note that if one of the two INI files exists there will be a flat namespace {yourTestNamespace}.INI, meaning that sections in the INI file are ignored. An example: if your test functions are hosted by a namespace Foo and your INI file specifies a variable hello as holding "world" then:
'world'≡#.Foo.INI.hello 1
Initialisation
Now the Run* method checks whether there is a function Initial in the hosting namespace. If that is the case then the function is executed.
Note that the function must be either niladic or monadic and must not return a result and, since version 3.0, may or may not return a Boolean result. A 1 means that function did what it is supposed to do (=same as no result) while a 0 means it could not initialize.
Of course you can simply execute → on a single line in your Initial function if any requirement is not met but that would also mean that your test did not run at all. If you run your test case automatically somehow than it should return a 1 indicating failure.
If the function is monadic then a Boolean is passed as the right argument telling the Initial function whether the test suite is going to run in batch mode (1) or not.
Use this function to create stuff that's needed by all test cases, or tell the user something important - of course only if the batch flag is false.
It might be a good idea to call a function tidying up in line 1, just in case a failing test case has left something behind; see below for details.
Another thing to mention is that Initial is not a bad place to establish the Helpers - see there for details.
Finally: running the test cases
Now the test cases are executed one by one.
Tidying up
After the last test case was executed the Run* function checks whether there is a function Cleanup in the namespace hosting your test cases. If that's the case then this function is executed. Such a function must be a niladic, no-result returning function.
INI file again
Finally the namespace "INI" holding variables populated from your INI file(s) is deleted.
Helpers
Over time it emerged that certain helpers are very useful in the namespace hosting the test cases. Therefore with version 1.4.0 a method EstablishHelpersIn was introduced that takes a reference as the right argument. That ref should point to the namespace hosting the test cases. If you are executing #.Testers.EstablishHelpersIn from within that namespace you can specify an empty vector as the right argument: it then defaults to ⎕IO⊃⎕RSI.
One of the helpers is the method ListHelpers which returns a table with all names in the first column and the first line of the function (which is expected to be a comment telling what the function actually does) in the second column:
ListHelpers Run Run all test cases RunDebug Run all test cases with DEBUG flag on RunThese Run just the specified tests. RunBatchTests Run all batch tests RunBatchTestsInDebugMode Run all batch tests in debug mode (no error trapping) and with stopFlag←1. E Get all functions into the editor starting their names with "Test_". L Prints a list with all test cases and the first comment line to the session. G Prints all groups to the session. FailsIf Usage : →FailsIf x, where x is a boolean scalar PassesIf Usage : →PassesIf x, where x is a boolean scalar GetTestFnsTemplate ⍝ Returns a vector of text vectors with the code of a template test case. GoToTidyUp Returns either an empty vector or "Label" which defaults to ∆TidyUp RenameTestFnsTo Renames a test function and tell acre. ListHelpers Lists all helpers available from the `Tester` class. ∆... ⍝ Niladic functions returning an integer constant, see table above.
The helpers fall into four groups:
Running test cases
Run, RunDebug, RunThese, RunBatchTest and RunBatchTestsInDebugMode are running all or selected test cases with or without error trapping. They are shortcuts for calling the methods in Tester.
Flow control
FailsIf, PassesIf and GoToTidyUp control the flow in test functions.
These functions return a result (Boolean) in case error trapping is active but make the calling Test* crash otherwise, allowing you to investigate a failing test case right on the spot.
This is achieved by the functions FailsIf and PassesIf signalling an error that can be trapped with:
⎕TRAP←(999 'C' '. ⍝ Deliberate error')(0 'N')
That's why the template for the test function carries such a statement and keeps ⎕TRAP local.
Note that GoToTidyUp allow you with a statement like
→GoToTidyUp expected≢result
to jump to a label ∆TidyUp. This is useful in case a test case needs to do some cleaning up.
Test function template
In case the namespace has no test functions in it (yet) the EstablishHelpersIn function also creates a test function template named Test_000. (Not when the hosting namespace is scripted!)
"Constants"
These niladic functions behave like constants and are useful for assigning explicit results in test functions.
Misc
E, G and L allow the user to edit all test functions and to list test groups, if any.
Examples
The test template contains examples for how to use the flow control functions.
The Run* functions are discussed later.
The most important helper functions are probably G and L: with a large number of test functions it's the only reasonable way to gather information: G lists all groups and L all test functions.
Since version 1.8.0 L requires a right argument (empty or group name) and an optional left argument (test case numbers).
Examples from the Fire development workspace:
⍴L '' 99 2 G acre InternalMethods List Misc Replace ReportGhosts Search L 'Li' Test_List_001 Unnamed NSs & GUI instances are ignored. Test_List_002 Just GUI instances are ignored Test_List_003 Nothing is ignored at all; instances of GUI objects are treaded like namespaces Test_List_004 Nothing is ignored but class instances Test_List_005 Scan all object types but with "Recursive"≡None Test_List_006 Scan all object types but with "Recursive"≡Just one level Test_List_007 Search for "Hello" with a ref and an unnamed namespace 1 7 99 L 'List' Test_List_001 Unnamed NSs & GUI instances are ignored. Test_List_007 Search for "Hello" with a ref and an unnamed namespace
Note that you don't need to specify more characters than needed to actually identify a group.
RunThese
A particularly helpful method while developing/enhancing stuff is RunThese. The function allows you to run just selected test functions rather than a whole test suite but at the same time process any INI files and execute Initial and Cleanup if they exist.
RunThese offers the following options:
RunThese 1 2 ⍝ Execute test cases 1 & 2 that do not belong to any group. RunThese ¯1 ¯2 ⍝ Same but stop just before a test case is actually executed. RunThese 'Group1' ⍝ Execute all test cases belonging to group 1. RunThese 'Group1' (2 3) ⍝ Execute test cases 2 & 3 of group 1. RunThese 'Group1' 2 3 ⍝ Same as before. RunThese 'L' ⍝ Execute all test cases of the group that start with "L" - if there is just one RunThese 'L*' ⍝ Execute all test cases of all groups starting with "L" RunThese 'Test_Group1_001' ⍝ Executes just this test case. RunThese 0 ⍝ Run all test cases but stop just before the execution.
Best Practices
- Try to keep your test cases simple and test just one thing at a time, for example just one method at a time. If a method is complex you will be better off creating a group (use the method name as group name) with several simple test cases rather than one complex test case.
- For more complex method create a group with the method name as group name.
- Create everything you need on the fly and tidy up afterwards. Or more precisely, tidy up (left overs!), prepare, test, tidy up again.
- Avoid a test case relying on changes made by an earlier test case. It's a tempting thing to do but you will almost certainly regret this at one stage or another.
- Notice that the DRY principle (don't repeat yourself) can and should be ignored in test cases: any test case should read from top to bottom like an independent story.
Tester's methods
Show the methods
Examples
Show the examples
Project Page
For bug reports, future enhancements and a full version history see Tester/ProjectPage
Version Information
Original author: |
|
Responsible: |
|
Email: |