local-it/mitgliederverwaltung
2
3
Fork 0
Code Issues 72 Pull requests 5 Projects 4 Releases Packages 1 Activity Actions

Implement fuzzy search #187

Merged
carla merged 10 commits from feature/162_fuzzy_search into main 2025-11-12 13:10:32 +01:00
Conversation 2 Commits 10 Files changed 6 +63
1 changed files with 63 additions and 0 deletions
Download patch file Download diff file Expand all files Collapse all files
Showing only changes of commit f6bfeadb7b - Show all commits

63
lib/membership/member.ex
View file

@ -3,6 +3,11 @@ defmodule Mv.Membership.Member do
domain: Mv.Membership, domain: Mv.Membership,
data_layer: AshPostgres.DataLayer data_layer: AshPostgres.DataLayer
require Ash.Query
import Ash.Expr
@default_fields [:first_name, :last_name, :email, :phone_number, :city, :street, :house_number, :postal_code]
postgres do postgres do
table "members" table "members"
repo Mv.Repo repo Mv.Repo
@ -108,6 +113,47 @@ defmodule Mv.Membership.Member do
where [changing(:user)] where [changing(:user)]
end end
end end
read :search do
argument :query, :string, allow_nil?: true
argument :fields, {:array, :atom}, allow_nil?: true
argument :similarity_threshold, :float, allow_nil?: true
prepare fn query, _ctx ->
q = Ash.Query.get_argument(query, :query) || ""
fields = Ash.Query.get_argument(query, :fields) || @default_fields
threshold = Ash.Query.get_argument(query, :similarity_threshold) || 0.2
carla marked this conversation as resolved
moritz commented 2025-11-05 17:46:03 +01:00
Copy link

some moduledocs would be nice for the public search function.

some moduledocs would be nice for the public `search` function.
if is_binary(q) and String.trim(q) != "" do
q2 = String.trim(q)
pat = "%" <> q2 <> "%"
# FTS as main filter and fuzzy search just fo first name, last name and strees
query
|> Ash.Query.filter(
carla marked this conversation as resolved
moritz commented 2025-11-05 17:32:20 +01:00
Copy link

the fields argument is never used in the search function.

the `fields` argument is never used in the `search` function.
expr(
fragment("search_vector @@ websearch_to_tsquery('simple', ?)", ^q2) or
fragment("search_vector @@ plainto_tsquery('simple', ?)", ^q2) or
# Substring on numeric-like fields (best effort, supports middle substrings)
contains(postal_code, ^q2) or
contains(house_number, ^q2) or
contains(phone_number, ^q2) or
fragment("? % first_name", ^q2) or
fragment("? % last_name", ^q2) or
fragment("? % street", ^q2) or
fragment("word_similarity(?, first_name) > ?", ^q2, ^threshold) or
fragment("word_similarity(?, last_name) > ?", ^q2, ^threshold) or
fragment("word_similarity(?, street) > ?", ^q2, ^threshold) or
fragment("similarity(first_name, ?) > ?", ^q2, ^threshold) or
fragment("similarity(last_name, ?) > ?", ^q2, ^threshold) or
fragment("similarity(street, ?) > ?", ^q2, ^threshold)
)
)
else
query
end
carla marked this conversation as resolved
moritz commented 2025-11-05 17:39:31 +01:00
Copy link

maybe the duplicate code for first name last name and street could be generated dynamically by the fields argument?

maybe the duplicate code for `first name` `last name` and `street` could be generated dynamically by the fields argument?
carla commented 2025-11-12 11:53:16 +01:00
Copy link

I decided to omit fields now...but maybe during refactoring we could think about a dynamic way

I decided to omit fields now...but maybe during refactoring we could think about a dynamic way
end
end
carla marked this conversation as resolved
moritz commented 2025-11-05 17:29:31 +01:00
Copy link

filter for email is missing
adding this worked for me:

                fragment("? % email", ^q2) or
                fragment("word_similarity(?, email) > ?", ^q2, ^threshold) or
                fragment("similarity(email, ?) > ?", ^q2, ^threshold) or
filter for email is missing adding this worked for me: ``` fragment("? % email", ^q2) or fragment("word_similarity(?, email) > ?", ^q2, ^threshold) or fragment("similarity(email, ?) > ?", ^q2, ^threshold) or ```
carla commented 2025-11-12 11:20:49 +01:00
Copy link

You're totally right, i forgot the email field. But I would favor for a simple contains ilike on the email field, because I think fuzzy search on the name already is enough? or what do you think?

You're totally right, i forgot the email field. But I would favor for a simple contains ilike on the email field, because I think fuzzy search on the name already is enough? or what do you think?
carla commented 2025-11-12 11:54:04 +01:00
Copy link

I added it as ilike and not as fuzzy search. But we can create another issue if we see we need it :)

I added it as ilike and not as fuzzy search. But we can create another issue if we see we need it :)
end end
validations do validations do
@ -281,4 +327,21 @@ defmodule Mv.Membership.Member do
identities do identities do
identity :unique_email, [:email] identity :unique_email, [:email]
end end
# Fuzzy Search function that can be called by live view and calls search action
def fuzzy_search(query, opts) do
q = (opts[:query] || opts["query"] || "") |> to_string()
if String.trim(q) == "" do
query
else
args =
case (opts[:fields] || opts["fields"]) do
nil -> %{query: q}
fields -> %{query: q, fields: fields}
end
Ash.Query.for_read(query, :search, args)
end
end
end end