Iaroslav Zeigerman
Iaroslav Zeigerman
@ChristopherDavenport, per our discussion on Gitter sharing my implementation of the SSE (Server-Sent Events) client: https://gist.github.com/izeigerman/09cf83a44a590e734873e552ae060fc4 This implementation supports all SSE attributes (like `id` and `retry`) as well as automatic...
Context from @StrikerRUS: >>> Now Go is failing (refer to https://github.com/BayesWitnesses/m2cgen/pull/200#issuecomment-624063683): ``` =================================== FAILURES =================================== _ test_e2e[xgboost_XGBClassifier - go_lang - train_model_classification_binary2] _ estimator = XGBClassifier(base_score=0.6, booster='gblinear', colsample_bylevel=None, colsample_bynode=None, colsamp...ambda=0, scale_pos_weight=1,...
Instead of including the whole vector clock into the Status message we should include a hash that represents the current version of data on a node. This would allow us...
In order to ensure a causality of messages when delivering them from one node to another we should maintain a separate message queue per each recipient. Each next message in...
Prevent the forward-only categorization of changes that require the creation of a new physical table
Currently there are no limitations on what can be changed as part of the forward-only plan / model. This includes model attributes that might require the creation of a new...
Sometimes a directly modified model may be associated with 100s indirectly modified downstream models. When this happens the output becomes cumbersome and overwhelming to a user. We need to revisit...
Currently SQLMesh always sets the `type` value to None when creating a relation instance returned by the `adapter.get_relation` and similar methods.
Right now if someone happens to run the SQLMesh state migration while there is an ongoing plan application, it may lead lead to unexpected results, lost data intervals and even...
Currently if an error occurs during snapshot table creation it immediately bubbles up, causing termination of the creation process. We should instead attempt to create tables for all snapshots for...