Create auto-generated objects.go file #178
Conversation
* ObjectType field of the IBObject is replaced with the ObjectType method for each child object. * Override ReturnFields method for each struct. By default, it will return a set of standard fields for each respective struct. * Removed Ipv4Addr/Ipv6Addr fields from the HostRecord object. * Updated existing object manager unit tests. * The IBObjectManager interface is not changed.
* Created ConnectorFacadeE2E that will delete objects created by test, when the test is done. * Created a set of e2e tests to check simple CRUD operations on the generated objects.go
* This way any change in the static part doesn't affect the generator workflow.
…fields * Previously used time.Time didn't unmarshalled right. Replaced it with a UnixTime struct which encapsulates the time.Time.
* Parsed WAPI doc pages are used for documentation comments by the generator.
* Test checks that such methods as ReturnFields, or ObjectType return valid values.
* Now ConnectorFacadeE2E will use slice of string instead of map for the deleteSet. This change is made to ensure the deletion order while sweeping objects. * Created a CRUD test for the Host Record Infoblox object.
* Tests mimic the qa-nios-api-tests.
…lient * Tests mimic the qa-nios-api-tests.
* Updated all unit and end-to-end tests to use v2 of the ginkgo BDD test suite.
* Now each end-to-end test has a set of labels, that describes if it's RO only, or RW test. Also added test IDs for wapi_framework_tests, those IDs are equal to the IDs in the original qa-nios-api-tests suite.
* Ginkgo v2 supports table driven tests, so objects_generated_test.go were rewritten using the ginkgo DescribeTable functionality.
* Tests mimic the qa-nios-api-tests.
* Now ConnectorFacadeE2E is able to update objects.
* Tests mimic the qa-nios-api-tests.
…er_merge # Conflicts: # objects.go
* Applied some tweaks to objects_generated.go. * Updated some tests with new assertions. * Created utils package, which can be used as a set of infoblox-go-client helper functions.
* Since updatable boolean field could be updated from true to false, we need to serialize false values, when updating some resource, instead of omit them.
* The tests listed all the ranges before, that could result in a wrong range object coming up first in the API server response. * Added query params to the test, so that it gets the right range object.
# Conflicts: # objects.go
There was a problem hiding this comment.
The major comments:
- 'omitempty' attribute will be harmfull in some cases; for example, if 'comment' field was set and a user wants to make it empty, it will not be possible to do using the implementation from this PR. As we discussed, this may be solved using a pointer to field's value instead of the value itself in the structure.
- It is hard to maintain the code located in a single file, at least when two or more people work on different issues//features but the code is in the same file. We can avoid this splitting the implementation into several files, one for each object type, for example.
Minor comment. It would be better to put an empty line between a field's definition in a structure and a comment for the next field in this structure. Just to improve readability.
…ointers
* Added empty lines between field definitions in objects_generated.go for
better readability.
* Changed updatable string and uint32 fields to pointers since they can be
updated to default values ("" and 0 respectively). Converting those
fields to pointers will make it possible for json lib to distinguish
between default and absent value.
|
Hello @skudriavtsev,
I couldn't think of a scenario, in which
This code is meant to be auto-generated. If somebody needs to change this file, this person should update the generator configuration first, and then use the generator to update the file. In this case, I don't think that it's gonna be hard to maintain the code, since there are no merge conflicts, when we have the generator which should always create the same code, if configuration isn't changed. Also it's much more easier to explain developers that they can't manually update one file, instead of articulating the same thought about the 100+ auto-generated files.
Added empty lines between field definitions. |
|
At least, 'comment' field is not fixed: updating with an empty comment leads to no update at all. At many objects. Besides, 'extattrs' field is also marked as 'omitempty', and this leads to the situation when we try to update with an empty list of EAs (removing all EAs) but update does not happen. This behavior must be changed. |
* Created e2e test case that checks if DNS View object's comment can be updated to empty string.
@skudriavtsev, could you provide the object with a problem Tested it out on the However you should know, that API will not return an empty string, when comment is set to empty string, even if you specify comment in |
|
Sorry, it seems that I have not retrieved the latest version of the code from your repository :-( I will re-test. |
* Updated generation logic for objects_generated.go. Now omitempty json tag is skipped for all extensible attribute fields. * Created e2e test for IPv4 Network object, which checks if all EAs of this object can be removed through update operation.
Changed EA behavior in 0e3d852. Now one can remove all extensible attributes via update operation. |
…records * Two new e2e tests validate if view field is updatable in HOST and CNAME record objects. * Fixed a small bug in ConnectorFacadeE2E that in some specific cases led to e2e test panic.
objects_generated.go
Outdated
|
|
||
| // This is a list of IPv6 Addresses for the host. | ||
| Ipv6Addrs []HostRecordIpv6Addr `json:"ipv6addrs,omitempty"` |
There was a problem hiding this comment.
@skudriavtsev what this fixes exactly?
This reverts commit c9a4cc5.
* After e2e testing of a client on a new env, a couple of assertion errors came up. Fixed those errors by changing the assertion rules.
* For environments with multiple resources search fields should be specified in some cases to get the needed resource.
* If view argument is empty in NewRecordCNAME, View field in RecordCNAME should be set to nil, to mimic old go-client behaviour.
* Added possibility to remove ip addresses from HostRecord object, by skipping omitempty for record:host_ipv4addr and record:host_ipv6addr WAPI types. * Added check for empty view string in the NewHostRecord method. * Added populateNilLists method for WapiRequestBuilder.
…_generation' into feature/e2e_testing_client_generation
Issue/Feature description
How it was fixed/implemented
ObjectType field of the IBObject is replaced with the ObjectType
method for each child object.
Override ReturnFields method for each struct. By default, it will return
a set of standard fields for each respective struct.
Removed Ipv4Addr/Ipv6Addr fields from the HostRecord object.
Updated existing object manager unit tests.
The IBObjectManager interface is not changed.
Tests
Created a E2E connector facade, that will remove all objects created by e2e test after the test is done.
Created a set of E2E tests to check crud operations on the generated object structs.
Reviewers
@skudriavtsev