Fix the NpgsqlCube to use the full G17 floating-point format - #6295
#6295Fix the NpgsqlCube to use the full G17 floating-point format#6295
Conversation
kirkbrauer
commented
Nov 6, 2025
- Switch cube floating-point string to use G17 format for 17 digits of precision to follow IEEE 754
|
@roji one question I did have is if we should remove the redundant @NinoFloris any thoughts on this? |
roji
left a comment
There was a problem hiding this comment.
one question I did have is if we should remove the redundant ToSubset() method in the NpgsqlCube now that we have the Subset() extension method.
I don't think the API of NpgsqlCube should be changed as a function of what we're doing in the EF provider - there may be Npgsql users which don't use EF and want to interact with Cube. So if we think that a client-side implementation of Subset makes sense to have, we should keep it in Npgsql.
Besides, given the conclusion we seem to have gotten to in npgsql/efcore.pg#3651 (comment)), do we even need an extension method in the EF provider? Why not just translate the ToSubset method on NpgsqlCube?
Unrelated: I noticed that NpgsqlCube.cs uses a namespace block instead of a statement, can you please change that in this PR as well to clean up the code? All of Npgsql uses namespace statements.
|
@roji Ok, I think we will unify on the |