Bind

Define an API controller like this:

def index(conn, _params) do
users = conn.query_string
|> Bind.query(User)
|> Repo.all()
render(conn, :index, result: users)
end

Now your endpoint supports all these queries out of the box:

GET /users?name[contains]=john&sort=-id&limit=25
GET /users?salary[gte]=50000&location[eq]=berlin
GET /users?joined_at[lt]=2024-01-01&status[neq]=disabled
GET /users?options.prompt[contains]=motorbike

Bind is a flexible and dynamic Ecto query builder, for retrieving data flexibly without writing custom queries for each use case.

Installation

Add bind to your list of dependencies in mix.exs:

def deps do
[
{:bind, "~> 0.6.0"}
]
end

API

Bind.query(schema, params)

Parameters:

Returns: An Ecto query.

Usage Example

Create Ecto query:

query = Bind.query(%{ "name[eq]" => "Alice", "age[gte]" => 30 }, MyApp.User)

Alternatively, with a query string:

query = Bind.query("?name[eq]=Alice&age[gte]=30", MyApp.User)

And finally run the query to get results from the database:

results = Repo.all(query)

Here's how it looks in a controller:

def index(conn, params) do
images = conn.query_string
|> Bind.query(MyApp.Media.Image)
|> MyApp.Repo.all()
render(conn, :index, result: images)
end

Error handling

case Bind.query(%{ "name[eq]" => "Alice", "age[gte]" => 30 }, MyApp.User) do
{:error, reason} ->
IO.puts("Error building query: #{reason}")
query ->
results = Repo.all(query)
end

Filtering

Examples:

%{"name[eq]" => "Alice", "age[gte]" => 30}
%{
"name[starts_with]" => "A",
"age[gte]" => 18,
"role[in]" => "superuser,admin,mod",
"is_active[true]" => "",
"last_login[nil]" => false
}

List of comparison operators supported:

For PostgreSQL JSONB columns, use dot notation to search within JSON fields:

%{"options.prompt[contains]" => "motorbike"}
%{"metadata.duration[eq]" => "5"}
%{"config.settings[starts_with]" => "prod"}

Examples in URLs:

GET /videos?options.prompt[contains]=motorbike
GET /users?preferences.theme[eq]=dark
GET /posts?metadata.tags[contains]=elixir

Supported JSONB operators:

Empty (JSONB arrays)

Check whether a JSONB array is empty. Works on a column whose value is an array directly, or on a nested key whose value is an array:

# Column value is an array, e.g. tags = [123, 23]
%{"tags[empty]" => "true"} # tags is []
%{"tags[empty]" => "false"} # tags has at least one element
# Nested key, e.g. options = %{"foobar" => []}
%{"options.foobar[empty]" => "true"}
%{"options.foobar[empty]" => "false"}

Examples in URLs:

GET /videos?tags[empty]=true
GET /videos?options.foobar[empty]=false

Sorting

Use the sort parameter to specify sorting order:

%{"sort" => "-age"} # Sort by age descending
%{"sort" => "age"} # Sort by age ascending

If nothing specified, sorts by ID field ascending.

Pagination

Example:

%{"limit" => 20, "start" => 100}

Example for descending pagination:

GET /users?sort=-created_at&-start=-100

Query String Support

In a typical Phoenix controller, you can simply pass conn.query_string and get Ecto query back:

query_string = conn.query_string
|> Bind.query(query_string, MyApp.User)
|> MyApp.Repo.all()

Transforming Query Parameters

You can transform filter values before query is built:

"user_id[eq]=123&team_id[eq]=456"
|> Bind.map(%{
user_id: fn id -> HashIds.decode(id) end,
team_id: fn id -> HashIds.decode(id) end
})
|> Bind.query(MyApp.User)
|> Repo.all()

Transform specific fields

Bind.map(params, %{
user_id: fn id -> HashIds.decode(id) end,
name: &String.upcase/1
})

Transform multiple fields with regex pattern:

Bind.map(params, %{
~r/_id$/i => fn id -> HashIds.decode(id) end
})

Note: Value transformation only applies to filter fields (e.g. [eq], [gte]), not to sort/limit/pagination params.

Safe Parameter Transformation

Use Bind.map_safe/2 when transformations might fail. It returns {:ok, mapped_params} on success or {:error, reason} on failure:

params
|> Bind.map_safe(%{
asset_id: fn hash -> HashIds.decode!(hash) end
})

map_safe/2 automatically handles Result tuples ({:ok, value} / {:error, reason}) returned by mapper functions:

# Mapper returns {:ok, value} - automatically unwrapped
params
|> Bind.map_safe(%{
asset_id: fn hash -> decode_id(hash) end # returns {:ok, id}
})
# => {:ok, %{"asset_id[eq]" => id}}
# Mapper returns {:error, reason} - propagated as error
params
|> Bind.map_safe(%{
asset_id: fn hash -> decode_id(hash) end # returns {:error, "invalid"}
})
# => {:error, {:transformation_failed, "invalid"}}
# Mix of return types works seamlessly
Bind.map_safe(params, %{
user_id: fn id -> decode(id) end, # {:ok, val} or {:error, reason}
team_id: fn id -> decode!(id) end, # val or raises
name: &String.upcase/1 # val directly
})

Empty value handling:

Empty values (nil or "") are automatically removed from the result if a mapper is defined for that field. Fields without mappers preserve empty values:

params = %{"user_id[eq]" => "", "name[eq]" => ""}
Bind.map_safe(params, %{
user_id: fn id -> decode!(id) end
})
# => {:ok, %{"name[eq]" => ""}}
# user_id[eq] is removed (has mapper + empty value)
# name[eq] is kept (no mapper)

Difference between map/2 and map_safe/2:

# map/2 - raises on error
Bind.map(params, %{id: &decode!/1})
# => raises if decode! fails
# map_safe/2 - returns error tuple
Bind.map_safe(params, %{id: &decode!/1})
# => {:ok, mapped} or {:error, {:transformation_failed, reason}}

Access Control with Filters

You can use filters to enforce access control and limit what users can query. Filters compose nicely with the query builder:

def index(conn, _params) do
my_posts = conn.query_string
# User can only see their own posts
|> Bind.filter(%{"user_id[eq]" => conn.assigns.current_user.id})
# That are active
|> Bind.filter(%{"active[true]" => true})
|> Bind.query(Post)
|> Repo.all()
render(conn, :index, posts: my_posts)
end

Join Queries

Query fields on associated tables using colon syntax: association:field[operator]

GET /users?address:city[eq]=berlin
GET /users?address:city[contains]=ber&sort=-id

Fields that you can run join query must be white-listed:

def index(conn, _params) do
users = conn.query_string
|> Bind.query(User, joins: [:address]) # allow join queries
|> Repo.all()
end

Virtual Fields

Expose filters that keep the field[op] notation but aren't columns — bind parses the param, your function writes the SQL. Useful when a filter is more than one join away:

def index(conn, _params) do
chats = conn.query_string
|> Bind.query(Chat, virtual: %{asset_id: &Chats.scope_touching_asset/2})
|> Repo.all()
end
# GET /chats?asset_id[eq]=123 calls scope_touching_asset(query, 123)
def scope_touching_asset(query, asset_id) do
chat_ids =
from m in ChatMessage,
join: r in NodeRun, on: r.chat_message_id == m.id,
join: v in AssetVersion, on: v.node_run_id == r.id,
where: v.asset_id == ^asset_id,
select: m.chat_id
from c in query, where: c.id in subquery(chat_ids)
end

A bare function answers eq only. Declare more constraints as a keyword list; in values arrive as a list (comma-split from the query string):

virtual: %{
asset_id: [
eq: &Chats.scope_touching_asset/2,
in: &Chats.scope_touching_any_asset/2
]
}

For PostgreSQL tsvector columns, use the search operator:

%{"bio[search]" => "engineer"}

Examples in URLs:

GET /users?keywords[search]=engineer

This generates a query using PostgreSQL's full-text search:

WHERE keywords @@ to_tsquery('simple', 'engineer:*')