feat: add user stream Websocket into C++ SDK

[yluoc]

Summary

• Adding Polymarket V2 user stream WS to SDK, which includes header file, source file, cmakelists, examples, and testcases.

Verification

• tests/test_user_stream.cpp
• examples/user_stream_example.cpp

Notes

New functionality passes all testcases and the live run against Polymarket's real user-channel server.

[yluoc]

Hello Sebastian, I am interested in contributing to your Polymarket V2 C++ SDK. What I add so far is a user-stream websocket connection, it passes with real polymarket creds. Looking forward to hear from you. @SebastianBoehler

[SebastianBoehler]

Hi Bill, thanks for reaching out by email and for opening this PR! Contributions are very welcome, and I appreciate you adding this one!

[SebastianBoehler]

Thanks again for the contribution. I reviewed 868e9c1 and some additional local probes exposed lifecycle cases that I think we should address before merging:

1. Reconciliation runs before the subscription is restored. In handle_stream_gap(), the callback runs synchronously. On reconnect, the transport invokes it before sending the authenticated subscription. A REST refresh inside that callback can therefore miss changes between the refresh and subscription activation. Polymarket does not replay missed events. Please arrange recovery so consumers reconcile after subscription restoration, while retaining immediate gap notification to invalidate stale state. A test should check that ordering.

2. A connection timeout leaves the socket running but disables event delivery. connect() deactivates the stream without stopping the transport. With a delayed handshake, I reproduced connect() returning false, followed by is_connected() becoming true while a valid order event was silently dropped. Please stop the transport when returning failure and add a delayed-handshake regression test.

3. Terminal authentication rejection does not end run(). The rejection handler disconnects the socket, but the blocking loop waits for stop(). I reproduced rejection with close code 1008 while run() remained blocked. I'd expect terminal rejection to end that loop. Please cover rejection both during run() and before entering it.

Two smaller points: please split the larger new files where practical; I aim for roughly 300 lines per file. Also, is there a supported producer for the additional topic/type/payload parsing path? If not, please keep the parser focused on the actual wire format.

Does this match what you observed in your live testing? If you intended different behavior, I'd like to hear your reasoning. Could you implement these changes and the regression tests? I can then review the update. My additional checks used local servers; I did not independently test real Polymarket credentials.

[yluoc]

@SebastianBoehler Thanks for replying, all issues you mentioned are fixed, I also split the gaint files smaller and remove the topic/payloaded path. All changes passes the testcases and pass against the real server. Let me know if there still exists any issues.

[yluoc]

Hello Sebastian, I fixed all issues you mentioned, I also splited the giant files into smaller one and remove the topic/payload parse path. All changes passes testcases and real server. Since Geographic Restriction I can't test or try the order-related SDK from USA, so maybe there still exists some problems I can't find out. If there are still exist any issues, please let me know. Moreover, if you want, we can use Discord and github PR comment to connect. Best, Bill Sebastian Boehler ***@***.***>

…

*SebastianBoehler* left a comment (SebastianBoehler/polymarket-cpp-client#25) <#25 (comment)> Thanks again for the contribution. I reviewed 868e9c1 and built the user-stream example and relevant test targets. All 14 selected tests passed. Some additional local probes exposed lifecycle cases that I think we should address before merging: 1. *Reconciliation runs before the subscription is restored.* In handle_stream_gap() <https://github.com/yluoc/polymarket-cpp-client/blob/868e9c1a34fd2262a2d2ca69a434b31eaf5b63aa/src/user_stream_runtime.cpp#L267-L274>, the callback runs synchronously. On reconnect, the transport invokes it before sending the authenticated subscription. A REST refresh inside that callback can therefore miss changes between the refresh and subscription activation. Polymarket does not replay missed events <https://docs.polymarket.com/trading/realtime-order-updates#recover-after-reconnecting>. Please arrange recovery so consumers reconcile after subscription restoration, while retaining immediate gap notification to invalidate stale state. A test should check that ordering. 2. *A connection timeout leaves the socket running but disables event delivery.* connect() <https://github.com/yluoc/polymarket-cpp-client/blob/868e9c1a34fd2262a2d2ca69a434b31eaf5b63aa/src/user_stream_runtime.cpp#L220-L230> deactivates the stream without stopping the transport. With a delayed handshake, I reproduced connect() returning false, followed by is_connected() becoming true while a valid order event was silently dropped. Please stop the transport when returning failure and add a delayed-handshake regression test. 3. *Terminal authentication rejection does not end run().* The rejection handler <https://github.com/yluoc/polymarket-cpp-client/blob/868e9c1a34fd2262a2d2ca69a434b31eaf5b63aa/src/user_stream_runtime.cpp#L277-L291> disconnects the socket, but the blocking loop waits for stop(). I reproduced rejection with close code 1008 while run() remained blocked. I'd expect terminal rejection to end that loop. Please cover rejection both during run() and before entering it. Two smaller points: please split the larger new files where practical; I aim for roughly 300 lines per file. Also, is there a supported producer for the additional topic/type/payload parsing path? If not, please keep the parser focused on the actual wire format. Does this match what you observed in your live testing? If you intended different behavior, I'd like to hear your reasoning. Could you implement these changes and the regression tests? I can then review the update. My additional checks used local servers; I did not independently test real Polymarket credentials. — Reply to this email directly, view it on GitHub <#25?email_source=notifications&email_token=AXTESC63DNRGPAN3V2PRTQT5R7OVPA5CNFSNUABFM5UWIORPF5TWS5BNNB2WEL2JONZXKZKDN5WW2ZLOOQXTKOJVG42DOMJSGEYKM4TFMFZW63VGMF2XI2DPOKSWK5TFNZ2KYZTPN52GK4S7MNWGSY3L#issuecomment-5957471210>, or unsubscribe <https://github.com/notifications/unsubscribe-auth/AXTESC674IQDZBR765BIWUT5R7OVPAVCNFSNUABGKJSXA33TNF2G64TZHMYTCMRWGY3TCNRUGQ5US43TOVSTWNJWGY4DMMZWGA2DTILWAI> . You are receiving this because you authored the thread.Message ID: ***@***.***>