Under QUICKSTART > Prerequisites the "Look Here" link returns a 404 error.

https://github.com/cerner/cclunit-framework/blob/master/doc/QUICKSTART.md#prerequisites

The problem is that some i4 variables and uar parameters have been declared as type h.

Currently, CCL resets the error stack whenever a child script is executed with the "execute" command. This could lead to errors being suppressed in a unit test if the error occurs before "execute" is invoked. An example of this would be if an error occurred while trying to create a mock table with cclutDefineMockTable and then the script-under-test is called using cclutExecuteProgramWithMocks.

Because the Mocking APIs provide an abstraction for "execute" now (cclutExecuteProgramWithMocks), the framework would be capable of checking for any errors in the error stack and documenting them, so that the test can fail even if all assertions pass (similar to what the cclutAssert* subroutines do). In fact, all of the Mocking APIs could support it to provide more information to the consumer about any errors in their unit test.

This issue is to implement this enhancement.

Currently, when using cclutAddMockIndex, the index will fail to be added to the table if the username of the user executing the test is greater than 7 characters in length when cclutCreateMockTable is called. The failure does generate an error, but if no cclutAssert* calls are made after cclutCreateMockTable but before execute/cclutExecuteProgramWithMocks, then the error will be suppressed (see here: #20)

The issue is similar to (#12) where there is a hard limit of 30 characters on Oracle identifiers, and CCL prepends XIE_ on to the name of any index (which also uses the tablename). Based on how the framework constructs the tablename, that leaves 7 characters for the username before hitting the limit.

Solutions would be similar for the constraint issue:

1. CCL could be enhanced to allow for specifying the index name when using "SELECT INTO table".
   or
2. The CCL Unit mocking framework could append indexes after the table has been finalized through the use of rdb commands (and thus specifying the specific index name to be used).

CCLUT should perform a preliminary compile of the test case like the maven plugin does to make it easier to identify syntax problems with the test case.

The doc page for assertions has a typo/spelling-mistake in the name of the assertion function:
cclutAssertDateTimeEqual: it is documented as 'cclutAssertDatetTimeEqual' (with an extra t)

took me some time to find this extra 't'....

page: https://github.com/cerner/cclunit-framework/blob/master/doc/CCLUTASSERTS.md

Currently, CCL's "SELECT INTO table" command offers an option to include constraints. However, there is no way to specify the name of the constraint. It will always be created as "CON__". Oracle requires that constraint names be no longer than 30 characters, which can be painful for custom tables with even semi-long table names or column names (dividing it evenly would leave only 13 characters for a table name and 12 for a column name). See an example here: https://connect.ucern.com/thread/255048 (Note that a uCern account is required.)

The CCL Unit mocking framework creates fairly long table names to ensure uniqueness when working in real domains, making this functionality near unusable. To resolve this, one of two things would need to happen.

1. CCL could be enhanced to allow for specifying the constraint name when using "SELECT INTO table".
   or
2. The CCL Unit mocking framework could append constraints after the table has been finalized through the use of rdb commands (and thus specifying the specific constraint name to be used).

Originally, (before understanding the limitation of constraint names), an implementation was developed using the constraint functionality of "SELECT INTO table". Its implementation is provided here as an example for any future work done on this enhancement:
https://gist.github.com/cernertrevor/acda50b4f6f60a4c0e35335158133602

Apparently the contents of _memory_reply_string trump generated report data in the vccl_run_program reply which is what DvDev displays.

A potential fix is to wipe out _memory_reply_string before generating the report, but there are pieces of the framework code that use _memory_reply_string to communicate error information back to DvDev. That communication mechanism would need adjusted to support this simple fix.

As a workaround one can add code to set _memory_reply_sting = " " in a teardownOnce routine.

When I try to create the target site, I am failing to create it because of some comments on the same line as a declare statement. When I remove the comments the problem is fixed.

Command run

mvn clean test site -Psoldev64aix

Line in question:

declare generatePortalInvites(dPatientPersonId = f8, dPmHistTrackingId = f8) = null ;569134 added dPmHistTrackingId

Error Message:

[INFO] --- maven-site-plugin:3.3:site (default-site) @ ccl-parent-pom ---
[INFO] configuring report plugin com.cerner.ccl.testing:cerreal-maven-plugin:2.1
[INFO] configuring report plugin com.cerner.ccl.whitenoise:whitenoise-maven-plugin:2.5
[INFO] configuring report plugin com.cerner.ccl.cdoc:cdoc-maven-plugin:1.2
[info] entering CDocReportMojo constructor
[INFO] configuring report plugin org.apache.maven.plugins:maven-project-info-reports-plugin:2.2
[INFO] configuring report plugin org.apache.maven.plugins:maven-site-plugin:3.3
[WARNING] No project URL defined - decoration links will not be relativized!
[INFO] Rendering site with org.apache.maven.skins:maven-default-skin:jar:1.0 skin.
[INFO] Generating "CCL Coverage Report" report --- cerreal-maven-plugin:2.1
[INFO] Generating "Cerreal Report" report --- cerreal-maven-plugin:2.1
[INFO] Generating "Whitenoise" report --- whitenoise-maven-plugin:2.5
[INFO] validating compile ccl_xml_translator_1557776900683.out
[INFO] Generating "CDoc" report --- cdoc-maven-plugin:1.2
[INFO] ------------------------------------------------------------------------
[INFO] BUILD FAILURE
[INFO] ------------------------------------------------------------------------
[INFO] Total time: 01:03 min
[INFO] Finished at: 2019-05-13T14:48:24-05:00
[INFO] ------------------------------------------------------------------------
[ERROR] Failed to execute goal org.apache.maven.plugins:maven-site-plugin:3.3:site (default-site) on project ccl-parent-pom: Error during page generation: Error rendering Maven report: Failed to parse
include file pfmt_pm_person_portal_invite.inc: Unrecognized declaration: null ;569134 added dPmHistTrackingId; normalized to NULL ;569134 ADDED DPMHISTTRACKINGID -> [Help 1]

CCL Unit Framework Version

3.3

Target folder

target.zip

This will happen if there is a variable in scope with the same name. Using a table alias will resolve the problem.

Builds were surviving for the wrong reason when cclut_find_unit_tests was failing to find any tests. That could not happen if testing were to fail for no tests. If there are no tests, why run the test command.

What if there are tests but no tests that match the specified test filters?

Due to the inactivity of this project for over 1 year, this project is marked to be archived at end of January 2024. Previously release assets will be untouched, but support will cease for this project going forward. If anyone has any reasons this project should not be archived, please comment.

So originally, this was proposed to be fixed with:
#19

But I don't think it resolves it in a meaningful way for a group1 tester. The code does work (it drops the definition when a group1 user is executing it and it goes down the tdbexecute route), but the problem is that the definition (while removed from dic.dat) is still in memory for the life of the CCL session. This means that any further attempts to overwrite the table (such as with a second test) will fail. Example below.

Script-under-test:

drop program group1test:dba go
create program group1test:dba

select into "nl:"
from person p
where p.person_id = 1
with nocounter

end
go

Test file

subroutine (tearDown(null) = null)
	call cclutRemoveAllMocks(null)
end ;tearDown

subroutine test_first(null)
	call cclutDefineMockTable("person", "person_id", "f8")
	call cclutCreateMockTable("person")
	call cclutAddMockData("person", "1.0")

	call cclutExecuteProgramWithMocks("group1test")
	call cclutAssertf8Equal(CURREF, "test_first 001", TRUE, TRUE)  
end

subroutine test_second(null)
	call cclutDefineMockTable("person", "person_id", "f8")
	call cclutCreateMockTable("person")
	call cclutAddMockData("person", "1.0")

	call cclutExecuteProgramWithMocks("group1test")
	call cclutAssertf8Equal(CURREF, "test_second 001", TRUE, TRUE)  
end

test_second will fail since it can't create the mocked person table (due to using the same names for mocks on new tests).

Given that CCL will still maintain that definition for the life of the session when it's dropped from a server, I don't know that there's any way to "resync" from CCL those definitions.

One possible alternative is to maintain a running counter for the unique index used when defining tables. Currently, it starts at 1 and counts up, based on how many existing tables there are. In the example above, since the tables are reset after each test, the mock person table is being created with the same name (CUST_CCLUT_1timestampUserid). If the counter was a running counter for the entire suite (such that test_second would create CUST_CCLUT_2timestampUserid), that would sidestep the issue. The tdbexecute logic would still stay in place because the tables still need to be dropped; this just allows group1 users to write more than one test leveraging mock tables.

So I would like to write tests for a piece of code which calls a service which in turn fetches data from two tables. Is there a way I can write tests using mocking api for something like this.