yuanbiao
/
emqx
1
0
Code Issues Pull Requests Packages Projects Releases Wiki Activity

feat(ds): ensure LTS state ids are deterministic

This commit is contained in:
Andrew Mayorov 2024-03-25 18:27:24 +01:00
parent 2cd357a5bd
commit d31cd0c728
No known key found for this signature in database
GPG Key ID: 2837C62ACFBFED5D
1 changed files with 4 additions and 2 deletions
Show Stats Download Patch File Download Diff File Expand all files Collapse all files

6
apps/emqx_durable_storage/src/emqx_ds_lts.erl
View File

@ -263,12 +263,14 @@ trie_insert(#trie{trie = Trie, stats = Stats, persist = Persist}, State, Token,
end. end.
-spec get_id_for_key(trie(), state(), edge()) -> static_key(). -spec get_id_for_key(trie(), state(), edge()) -> static_key().
get_id_for_key(#trie{static_key_size = Size}, _State, _Token) -> get_id_for_key(#trie{static_key_size = Size}, State, Token) when Size =< 32 ->
%% Requirements for the return value: %% Requirements for the return value:
%% %%
%% It should be globally unique for the `{State, Token}` pair. Other %% It should be globally unique for the `{State, Token}` pair. Other
%% than that, there's no requirements. The return value doesn't even %% than that, there's no requirements. The return value doesn't even
%% have to be deterministic, since the states are saved in the trie. %% have to be deterministic, since the states are saved in the trie.
%% Yet, it helps a lot if it is, so that applying the same sequence
%% of topics to different tries will result in the same trie state.
%% %%
%% The generated value becomes the ID of the topic in the durable %% The generated value becomes the ID of the topic in the durable
%% storage. Its size should be relatively small to reduce the %% storage. Its size should be relatively small to reduce the
@ -277,7 +279,7 @@ get_id_for_key(#trie{static_key_size = Size}, _State, _Token) ->
%% If we want to impress computer science crowd, sorry, I mean to %% If we want to impress computer science crowd, sorry, I mean to
%% minimize storage requirements, we can even employ Huffman coding %% minimize storage requirements, we can even employ Huffman coding
%% based on the frequency of messages. %% based on the frequency of messages.
<<Int:(Size * 8)>> = crypto:strong_rand_bytes(Size), <<Int:(Size * 8), _/bytes>> = crypto:hash(sha256, term_to_binary([State | Token])),
Int. Int.
%% erlfmt-ignore %% erlfmt-ignore