feat: Implement streaming data source. #179
Conversation
…dks-private into rlamb/data-store-arch
…arkly/cpp-sdks-private into rlamb/basic-store-implementation
kinyoklion
changed the title
| } | ||
| return DataSourceEventHandler::Patch{ | ||
| TStreamingDataKind::Key(path), | ||
| data_model::ItemDescriptor<TData>(data->value())}; |
There was a problem hiding this comment.
I wish this lint worked, because it would be useful, but it is basically always wrong. Checked on line 41.
kinyoklion
force-pushed
the
rlamb/implement-server-streaming-data-source
branch
from
2ca520e to
77de4fd
Compare
| return; | ||
| } | ||
|
|
||
| // TODO: Initial reconnect delay. sc-204393 |
There was a problem hiding this comment.
Likely this should still be addressed as an independent change, but maybe this ticket should be pulled into the epic.
I think that the SSE client currently has a hardcoded value and it just needs to expose a value for building instead, and then we can use that value.
| if (auto self = weak_self.lock()) { | ||
| self->data_source_handler_.HandleMessage(event.type(), | ||
| event.data()); | ||
| // TODO: Use the result of handle message to restart the |
There was a problem hiding this comment.
I am not sure how to handle this in the current pattern. I think we went with re-creating instead of restarting. So it will require extracting out the build of the client, then closing and re-creating the client.
There was a problem hiding this comment.
Another thought is have the callback return a bool/status that tells the eventsource to reconnect. Then we don't have to rebuild it and can keep the async stuff hidden there.
There was a problem hiding this comment.
That sounds more pleasant.
libs/internal/include/launchdarkly/data_sources/data_source_status_manager_base.hpp
Outdated
Show resolved
Hide resolved
libs/internal/include/launchdarkly/data_sources/data_source_status_manager_base.hpp
Outdated
Show resolved
Hide resolved
| PARSE_FIELD(path, obj, "path"); | ||
|
|
||
| if (StreamingDataKinds::Flag::IsKind(path)) { | ||
| return Patch<StreamingDataKinds::Flag, data_model::Flag>(path, obj); |
There was a problem hiding this comment.
I spent an inordinate amount of time on template shenanigans. Ultimately it is pretty pleasant though,
| LD_LOG(logger_, LogLevel::kWarn) | ||
| << "Polling for feature flag updates failed: " | ||
| << (error_message.has_value() ? *error_message : "unknown error"); | ||
| } else if (res.Status() == 200) { |
There was a problem hiding this comment.
Kinda wish we could use the beast error codes here. Maybe we can, they might just be the numeric values of the status codes or have a getter for it.
| if (auto self = weak_self.lock()) { | ||
| self->data_source_handler_.HandleMessage(event.type(), | ||
| event.data()); | ||
| // TODO: Use the result of handle message to restart the |
There was a problem hiding this comment.
Another thought is have the callback return a bool/status that tells the eventsource to reconnect. Then we don't have to rebuild it and can keep the async stuff hidden there.
…atus_manager_base.hpp Co-authored-by: Casey Waldren <cwaldren@launchdarkly.com>
Co-authored-by: Casey Waldren <cwaldren@launchdarkly.com>
Co-authored-by: Casey Waldren <cwaldren@launchdarkly.com>
Co-authored-by: Casey Waldren <cwaldren@launchdarkly.com>
|
|
||
| #include <launchdarkly/encoding/base_64.hpp> | ||
| #include <launchdarkly/serialization/json_flag.hpp> | ||
| // #include <launchdarkly/serialization/json_primitives.hpp> |
No description provided.