feat: Implement polling data source. #28
Conversation
|
|
||
| Client client(ConfigBuilder(key).Build().value(), | ||
| ContextBuilder().kind("user", "ryan").build()); | ||
| Client client( |
There was a problem hiding this comment.
Made this use polling for now. So I could watch it poll.
| * will be used. | ||
| */ | ||
| class StreamingDataHandler { | ||
| class DataSourceEventHandler { |
There was a problem hiding this comment.
Used by both polling and streaming.
|
|
||
| using launchdarkly::client_side::data_sources::DataSourceStatus; | ||
|
|
||
| static std::unique_ptr<IDataSource> MakeDataSource( |
There was a problem hiding this comment.
Once we have file and test data sources we may need to scale this differently.
| std::lock_guard lock(status_mutex_); | ||
|
|
||
| // If initializing, then interruptions remain initializing. | ||
| auto new_state = |
There was a problem hiding this comment.
I missed this bit of logic before.
| << (retry_message ? *retry_message : "giving up permanently"); | ||
| } | ||
|
|
||
| bool IsInvalidSdkKeyStatus(HttpResult::StatusCode code) { |
There was a problem hiding this comment.
I should maybe find a different home for this?
| message, std::chrono::system_clock::now()); | ||
| } | ||
|
|
||
| data_source_status_signal_(std::move(Status())); |
There was a problem hiding this comment.
encountering the same in my event processor stuff. I figure, I'll accept a std::function and then the caller can dispatch it to some strand using asio::post
apps/hello-cpp/main.cpp
Outdated
| // Sit around. | ||
| std::string t; | ||
| std::cout << "Press enter to exit" << std::endl; | ||
| std::cin >> t; |
There was a problem hiding this comment.
There may be some problem actually closing. So I need to look at that.
There was a problem hiding this comment.
I left it overnight and then it just didn't exit.
| Context const& context, | ||
| config::detail::built::ServiceEndpoints const& endpoints, | ||
| config::detail::built::HttpProperties const& http_properties, | ||
| std::chrono::seconds polling_interval, |
There was a problem hiding this comment.
I've been kind of debating this on my end, but wondering if we should just pass in the Config to keep the number of parameters down.
After all, it (should) be very easy to build a fully valid config with just a small delta for your use/test case.
That'd eliminate the sdk_key, endpoints, http_properties, and polling_interval
There was a problem hiding this comment.
Streaming and polling config have been re-worked with this and the defaults.
| inline const static std::string polling_get_path_ = "/msdk/evalx/contexts"; | ||
|
|
||
| inline const static std::string polling_report_path_ = | ||
| "/msdk/evalx/context"; |
There was a problem hiding this comment.
Might as well pull these into the PollingConfig?
There was a problem hiding this comment.
I still don't feel like these are config. You can never change them correct? Maybe I just need to make a place for constants.
There was a problem hiding this comment.
I guess if I think of the possibility of sharing the implementation, then possibly they could be configurable. In the sense that they may differ between implementations.
There was a problem hiding this comment.
I believe they are config even if they aren't user-facing, since they are config for us. I think it's nice to be able to see all of those variables in one place, like in Defaults (plus makes testing easier)
| message, std::chrono::system_clock::now()); | ||
| } | ||
|
|
||
| data_source_status_signal_(std::move(Status())); |
There was a problem hiding this comment.
encountering the same in my event processor stuff. I figure, I'll accept a std::function and then the caller can dispatch it to some strand using asio::post
|
|
||
| namespace launchdarkly::client_side::data_sources::detail { | ||
|
|
||
| const static std::chrono::seconds kMinPollingInterval = |
There was a problem hiding this comment.
Could pull into PollingConfig.
| // and it didn't have an etag, we still don't want to try to | ||
| // parse the body. | ||
| } else { | ||
| if (network::detail::IsRecoverableStatus(res.Status())) { |
There was a problem hiding this comment.
Ah, I wrote one of these too. We can compare once you merge this PR, maybe I can delete mine..
No description provided.