perf(import): reuse auto-created groups across import chunks
This commit is contained in:
parent
68a1a9530a
commit
118b9f8d57
5 changed files with 90 additions and 3 deletions
|
|
@ -3,6 +3,24 @@ defmodule Mv.Membership.Import.ImportRunnerTest do
|
|||
|
||||
alias Mv.Membership.Import.ImportRunner
|
||||
|
||||
describe "carry_groups_forward/2" do
|
||||
test "replaces import_state groups_found with the chunk's grown snapshot" do
|
||||
import_state = %{groups_found: [%{id: "1", name: "A"}]}
|
||||
chunk_result = %{groups_found: [%{id: "1", name: "A"}, %{id: "2", name: "B"}]}
|
||||
|
||||
assert ImportRunner.carry_groups_forward(import_state, chunk_result) == %{
|
||||
groups_found: [%{id: "1", name: "A"}, %{id: "2", name: "B"}]
|
||||
}
|
||||
end
|
||||
|
||||
test "leaves import_state unchanged when the chunk result omits groups_found" do
|
||||
import_state = %{groups_found: [%{id: "1", name: "A"}], other: :kept}
|
||||
chunk_result = %{inserted: 1}
|
||||
|
||||
assert ImportRunner.carry_groups_forward(import_state, chunk_result) == import_state
|
||||
end
|
||||
end
|
||||
|
||||
describe "read_file_entry/2" do
|
||||
test "returns {:ok, content} for a readable file" do
|
||||
path =
|
||||
|
|
|
|||
|
|
@ -1116,6 +1116,53 @@ defmodule Mv.Membership.Import.MemberCSVTest do
|
|||
"Expected the group table read at most a few times, got #{reads} reads for 10 rows (N+1)."
|
||||
end
|
||||
|
||||
test "returns the grown group snapshot so later chunks skip the table read",
|
||||
%{actor: actor} do
|
||||
chunk1 = [
|
||||
{2, %{member: %{email: "g-xchunk-1@example.com"}, custom: %{}, groups: "Shared X"}}
|
||||
]
|
||||
|
||||
chunk2 = [
|
||||
{3, %{member: %{email: "g-xchunk-2@example.com"}, custom: %{}, groups: "Shared X"}}
|
||||
]
|
||||
|
||||
assert {:ok, result1} =
|
||||
MemberCSV.process_chunk(chunk1, %{email: 0}, %{}, actor: actor, groups_found: [])
|
||||
|
||||
# The chunk result must expose the accumulated snapshot, including the group
|
||||
# auto-created while processing this chunk, so the LiveView can thread it
|
||||
# into the next chunk's opts.
|
||||
assert is_list(result1.groups_found)
|
||||
assert Enum.any?(result1.groups_found, &(&1.name == "Shared X"))
|
||||
|
||||
group_read_count = Agent.start_link(fn -> 0 end) |> elem(1)
|
||||
test_pid = self()
|
||||
|
||||
handler = fn _event, _measurements, metadata, _config ->
|
||||
if self() == test_pid and metadata[:source] == "groups" and
|
||||
is_binary(metadata[:query]) and String.starts_with?(metadata.query, "SELECT") do
|
||||
Agent.update(group_read_count, &(&1 + 1))
|
||||
end
|
||||
end
|
||||
|
||||
handler_id = "test-xchunk-group-read-#{System.unique_integer([:positive])}"
|
||||
:telemetry.attach(handler_id, [:mv, :repo, :query], handler, nil)
|
||||
|
||||
assert {:ok, %{inserted: 1}} =
|
||||
MemberCSV.process_chunk(chunk2, %{email: 0}, %{},
|
||||
actor: actor,
|
||||
groups_found: result1.groups_found
|
||||
)
|
||||
|
||||
reads = Agent.get(group_read_count, & &1)
|
||||
:telemetry.detach(handler_id)
|
||||
|
||||
# The second chunk receives the snapshot grown by the first, so the shared
|
||||
# group resolves from memory without any full-table read.
|
||||
assert reads == 0,
|
||||
"Expected no group table read in the second chunk, got #{reads} (snapshot not threaded across chunks)."
|
||||
end
|
||||
|
||||
test "empty groups cell leaves the member without group assignment", %{actor: actor} do
|
||||
chunk = [{2, %{member: %{email: "g-empty@example.com"}, custom: %{}, groups: " "}}]
|
||||
opts = [actor: actor, groups_found: []]
|
||||
|
|
|
|||
Loading…
Add table
Add a link
Reference in a new issue