make api_service method work

[eatwithforks]

@cben @moolitayer

[cben]

Thanks!

Let's see all resources with consecutive capital letters:

github.com/cben/kubernetes-discovery-samples$ ag -o '"kind":.*[A-Z][A-Z]\S*' --nonumbers --nofilename */api*/ | sort -u

"kind": "APIGroup",
"kind": "APIGroupList",
"kind": "APIResourceList",
"kind": "APIService",
"kind": "APIVersions",
"kind": "OAuthAccessToken",
"kind": "OAuthAuthorizeToken",
"kind": "OAuthClient",
"kind": "OAuthClientAuthorization",

(APIGroup, APIGroupList, APIResourceList are not actual API resources, they're kinds of the discovery results themselves)

kind	old method names	new method names
APIService	apiservice	api_service
APIVersions	apiversions	api_versions
OAuthAccessToken	oauth_access_token	o_auth_access_token
OAuthAuthorizeToken	oauth_authorize_token	o_auth_authorize_token
OAuthClient	oauth_client	o_auth_client
OAuthClientAuthorization	oauth_client_authorization	o_auth_client_authorization

and couple more example, one made up, one from istio CRD referenced in #361:

kind	old method names	new method names
lowerCamelUPPERCase	lower_camel_uppercase	lower_camel_upper_case
HTTPAPISpecBinding	httpapispec_binding	httpapi_spec_binding

Hmm, is this a breaking change? I agree it's better (and matches Rails' .underscore behaviour) 👍, but perhaps someone already has working code with say client.get_apiservice and now will need to change to client.get_api_service...
BTW, many of these kinds are recent but some go back quite a while, APIService appeared in kube 1.7...

• I definitely don't want to release this in a patch bump.
• compromise on calling this a bug but releasing in a minor bump?
• major bump?

[grosser]

I'd say either major or we make both available and call it a patch. Is there a planned major bump coming up that we can piggy back on?

…

On Tue, Oct 9, 2018 at 9:25 AM Beni Cherniavsky-Paskin < ***@***.***> wrote: Thanks! Let's see all resources with consecutive capital letters: github.com/cben/kubernetes-discovery-samples> ag -o 'kind:.*[A-Z][A-Z]\S*' --nonumbers --nofilename | sort | uniq kind: APIGroup kind: APIGroupList kind: APIResourceList kind: APIService kind: APIServiceList kind: APIVersions kind: OAuthAccessToken kind: OAuthAccessTokenList kind: OAuthAuthorizeToken kind: OAuthAuthorizeTokenList kind: OAuthClient kind: OAuthClientAuthorization kind: OAuthClientAuthorizationList kind: OAuthClientList kind old method names new method names APIGroup apigroup api_group APIGroupList apigroup_list api_group_list APIResourceList apiresource_list api_resource_list APIService apiservice api_service APIServiceList apiservice_list api_service_list APIVersions apiversions api_versions OAuthAccessToken oauth_access_token o_auth_access_token OAuthAccessTokenList oauth_access_token_list o_auth_access_token_list OAuthAuthorizeToken oauth_authorize_token o_auth_authorize_token OAuthAuthorizeTokenList oauth_authorize_token_list o_auth_authorize_token_list OAuthClient oauth_client o_auth_client OAuthClientAuthorization oauth_client_authorization o_auth_client_authorization OAuthClientAuthorizationList oauth_client_authorization_list o_auth_client_authorization_list OAuthClientList oauth_client_list o_auth_client_list Hmm, is this a breaking change? I agree it's better (and matches Rails' .underscore behaviour) 👍, but perhaps someone already has working code with say client.get_apiservice and now will need to change to client.get_api_service... BTW, many of these kinds are recent but some go back quite a while, APIService appeared in kube 1.7... - I definitely don't want to release this in a patch bump. - compromise on calling this a bug but releasing in a minor bump? - major bump? — You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub <#355 (comment)>, or mute the thread <https://github.com/notifications/unsubscribe-auth/AAAsZ1n1NX-pNk67pZKcptds8URmckA9ks5ujM3kgaJpZM4XN2AO> .

[cben]

I'm working on making both old and new work...

[cben]

Rebased, still blocked on compatibility, working on it.
I'm going to first rewrite parse_definition and add more tests in separate PR and then return to this...

[cben]

Rebased again.
But with recent PRs inflating the whole (kind,name)->method names derivation a lot, I'm not sure I want to complicate it further by computing computing both old and new names for compatibility :-(

[cben]

reminder to self: there will never be a perfect time but fixing this sooner is better.
I should merge this and bump major version sometime soon...

[cben]

Oh, wow. Sorry for holding this up so long.
I'm planning a major version bump soon, see #435