Current section

Files

Jump to
lotus lib lotus.ex
Raw

lib/lotus.ex

defmodule Lotus do
@moduledoc """
Lotus is a lightweight Elixir library for saving and executing read-only SQL queries.
This module provides the main public API, orchestrating between:
- Storage: Query persistence and management
- Runner: SQL execution with safety checks
- Migrations: Database schema management
## Configuration
Add to your config:
config :lotus,
repo: MyApp.Repo,
primary_key_type: :id, # or :binary_id
foreign_key_type: :id # or :binary_id
## Usage
# Create and save a query with variables
{:ok, query} = Lotus.create_query(%{
name: "Active Users",
statement: "SELECT * FROM users WHERE active = {{is_active}}",
variables: [
%{name: "is_active", type: :text, label: "Is Active", default: "true"}
],
search_path: "reporting, public"
})
# Execute a saved query
{:ok, results} = Lotus.run_query(query)
# Execute SQL directly (read-only)
{:ok, results} = Lotus.run_sql("SELECT * FROM products WHERE price > $1", [100])
"""
@default_page_size 1000
@type cache_opt ::
:bypass
| :refresh
| {:ttl_ms, non_neg_integer()}
| {:profile, atom()}
| {:tags, [binary()]}
@type window_count_mode :: :none | :exact
@type window_opts :: [
limit: pos_integer(),
offset: non_neg_integer(),
count: window_count_mode
]
@type opts :: [
timeout: non_neg_integer(),
statement_timeout_ms: non_neg_integer(),
read_only: boolean(),
search_path: binary() | nil,
repo: binary() | nil,
vars: map(),
cache: [cache_opt] | :bypass | :refresh | nil,
window: window_opts
]
alias Lotus.{Config, Storage, Runner, Result, Schema, Sources}
alias Lotus.Storage.Query
def child_spec(opts), do: Lotus.Supervisor.child_spec(opts)
def start_link(opts \\ []), do: Lotus.Supervisor.start_link(opts)
@doc """
Returns the current version of Lotus.
"""
def version do
Application.spec(:lotus, :vsn)
|> to_string()
end
@doc """
Returns the configured Ecto repository where Lotus stores query definitions.
"""
def repo, do: Config.repo!()
@doc """
Returns all configured data repositories.
"""
def data_repos, do: Config.data_repos()
@doc """
Gets a data repository by name.
Raises if the repository is not configured.
"""
def get_data_repo!(name), do: Config.get_data_repo!(name)
@doc """
Lists the names of all configured data repositories.
Useful for building UI dropdowns.
"""
def list_data_repo_names, do: Config.list_data_repo_names()
@doc """
Returns the default data repository as a {name, module} tuple.
- If there's only one data repo configured, returns it
- If multiple repos are configured and default_repo is set, returns that repo
- If multiple repos are configured without default_repo, raises an error
- If no data repos are configured, raises an error
"""
def default_data_repo, do: Config.default_data_repo()
@doc """
Lists all saved queries.
"""
defdelegate list_queries(), to: Storage
@doc """
Gets a single query by ID. Returns nil if not found.
"""
defdelegate get_query(id), to: Storage
@doc """
Gets a single query by ID. Raises if not found.
"""
defdelegate get_query!(id), to: Storage
@doc """
Creates a new saved query.
"""
defdelegate create_query(attrs), to: Storage
@doc """
Updates an existing query.
"""
defdelegate update_query(query, attrs), to: Storage
@doc """
Deletes a saved query.
"""
defdelegate delete_query(query), to: Storage
@doc """
Run a saved query (by struct or id).
Variables in the query statement (using `{{variable_name}}` syntax) are
substituted with values from the query's default variables and any runtime
overrides provided via the `vars` option.
## Variable Resolution
Variables are resolved in this order:
1. Runtime values from `vars` option (highest priority)
2. Default values from the query's variable definitions
3. If neither exists, raises an error for missing required variable
## Examples
# Run query with default variable values
Lotus.run_query(query)
# Override variables at runtime
Lotus.run_query(query, vars: %{"min_age" => 25, "status" => "active"})
# Run with timeout and repo options
Lotus.run_query(query, timeout: 10_000, repo: MyApp.DataRepo)
# Run by query ID
Lotus.run_query(query_id, vars: %{"user_id" => 123})
## Variable Types
Variables are automatically cast based on their type definition:
- `:text` - Used as-is (strings)
- `:number` - Cast from string to integer
- `:date` - Cast from ISO8601 string to Date struct
### Windowed pagination
Pass `window: [limit: pos_integer, offset: non_neg_integer, count: :none | :exact]` to
return only a page of rows from the original query. When `count: :exact`, Lotus will
also compute `SELECT COUNT(*) FROM (original_sql)` and include `meta.total_count` in
the result. The `num_rows` field always reflects the number of rows in the returned page.
"""
@spec run_query(Query.t() | term(), opts()) :: {:ok, Result.t()} | {:error, term()}
def run_query(query_or_id, opts \\ [])
def run_query(%Query{} = q, opts) do
supplied_vars = Keyword.get(opts, :vars, %{}) || %{}
defaults =
q.variables
|> Enum.filter(& &1.default)
|> Map.new(fn v -> {v.name, v.default} end)
vars = Map.merge(defaults, supplied_vars)
{sql, params} =
try do
Query.to_sql_params(q, vars)
rescue
e in ArgumentError ->
{:error, e.message}
end
case {sql, params} do
{:error, msg} ->
{:error, msg}
{sql, params} ->
{repo_mod, repo_name} = Sources.resolve!(Keyword.get(opts, :repo), q.data_repo)
search_path = Keyword.get(opts, :search_path) || q.search_path
final_opts = if search_path, do: Keyword.put(opts, :search_path, search_path), else: opts
{sql, params, window_meta, cache_bound} =
maybe_apply_window(
sql,
params,
repo_mod || repo_name,
search_path,
Keyword.get(opts, :window)
)
key = result_key(sql, cache_bound || vars, repo_name, search_path)
tags =
["query:#{q.id}", "repo:#{repo_name}"] ++
if is_list(opts[:cache]), do: Keyword.get(opts[:cache], :tags, []), else: []
profile =
if is_list(opts[:cache]) do
Keyword.get(opts[:cache], :profile, Config.default_cache_profile())
else
Config.default_cache_profile()
end
exec_with_cache(opts[:cache], profile, key, tags, fn ->
with {:ok, %Result{} = res} <- Runner.run_sql(repo_mod, sql, params, final_opts) do
{:ok, merge_window_meta(res, window_meta)}
end
end)
end
end
def run_query(id, opts) do
q = Storage.get_query!(id)
run_query(q, opts)
end
@doc """
Checks if a query can be run with the provided variables.
Returns true if all required variables have values (either from defaults
or supplied vars), false otherwise.
## Examples
# Query with all required variables having defaults
Lotus.can_run?(query)
# => true
# Query missing required variables
Lotus.can_run?(query)
# => false
# Query with runtime variable overrides
Lotus.can_run?(query, vars: %{"user_id" => 123})
# => true (if user_id was the missing variable)
"""
@spec can_run?(Query.t()) :: boolean()
@spec can_run?(Query.t(), opts()) :: boolean()
def can_run?(query, opts \\ [])
def can_run?(%Query{} = q, opts) do
supplied_vars = Keyword.get(opts, :vars, %{}) || %{}
defaults =
q.variables
|> Enum.filter(& &1.default)
|> Map.new(fn v -> {v.name, v.default} end)
vars = Map.merge(defaults, supplied_vars)
try do
Query.to_sql_params(q, vars)
true
rescue
ArgumentError ->
false
end
end
@doc """
Run ad-hoc SQL (bypassing storage), still read-only and sandboxed.
## Examples
# Run against default configured repo
{:ok, result} = Lotus.run_sql("SELECT * FROM users")
# Run against specific repo
{:ok, result} = Lotus.run_sql("SELECT * FROM products", [], repo: MyApp.DataRepo)
# With parameters
{:ok, result} = Lotus.run_sql("SELECT * FROM users WHERE id = $1", [123])
# With search_path for schema resolution
{:ok, result} = Lotus.run_sql("SELECT * FROM users", [], search_path: "reporting, public")
### Windowed pagination
Pass `window: [limit: pos_integer, offset: non_neg_integer, count: :none | :exact]` to
page results from the SQL. See `run_query/2` for details. The cache key automatically
incorporates the window so different pages are cached independently.
"""
@spec run_sql(binary(), list(any()), [
{:read_only, boolean()}
| {:statement_timeout_ms, non_neg_integer()}
| {:timeout, non_neg_integer()}
| {:search_path, binary() | nil}
| {:repo, atom() | binary()}
| {:window, window_opts}
]) ::
{:ok, Result.t()} | {:error, term()}
def run_sql(sql, params \\ [], opts \\ []) do
{repo_mod, repo_name} = Sources.resolve!(Keyword.get(opts, :repo), nil)
runner_opts = Keyword.delete(opts, :repo)
search_path = Keyword.get(runner_opts, :search_path)
{sql, params, window_meta, cache_bound} =
maybe_apply_window(
sql,
params,
repo_mod || repo_name,
search_path,
Keyword.get(opts, :window)
)
key = result_key(sql, cache_bound || params, repo_name, search_path)
tags =
["repo:#{repo_name}"] ++
if is_list(opts[:cache]), do: Keyword.get(opts[:cache], :tags, []), else: []
profile =
if is_list(opts[:cache]) do
Keyword.get(opts[:cache], :profile, Config.default_cache_profile())
else
Config.default_cache_profile()
end
exec_with_cache(opts[:cache], profile, key, tags, fn ->
with {:ok, %Result{} = res} <- Runner.run_sql(repo_mod, sql, params, runner_opts) do
{:ok, merge_window_meta(res, window_meta)}
end
end)
end
@doc """
Returns whether unique query names are enforced.
"""
defdelegate unique_names?(), to: Config
@doc """
Lists all tables in a data repository.
For databases with schemas (PostgreSQL), returns {schema, table} tuples.
For databases without schemas (SQLite), returns just table names as strings.
## Examples
{:ok, tables} = Lotus.list_tables("postgres")
# Returns [{"public", "users"}, {"public", "posts"}, ...]
{:ok, tables} = Lotus.list_tables("postgres", search_path: "reporting, public")
# Returns [{"reporting", "customers"}, {"public", "users"}, ...]
{:ok, tables} = Lotus.list_tables("sqlite")
# Returns ["products", "orders", "order_items"]
"""
def list_tables(repo_or_name, opts \\ []), do: Schema.list_tables(repo_or_name, opts)
@doc """
Lists all schemas in the given repository.
Returns a list of schema names. For databases without schemas (like SQLite),
returns an empty list.
## Examples
{:ok, schemas} = Lotus.list_schemas("postgres")
# Returns ["public", "reporting", ...]
{:ok, schemas} = Lotus.list_schemas("sqlite")
# Returns []
"""
def list_schemas(repo_or_name, opts \\ []), do: Schema.list_schemas(repo_or_name, opts)
@doc """
Gets the schema for a specific table.
## Examples
{:ok, schema} = Lotus.get_table_schema("primary", "users")
{:ok, schema} = Lotus.get_table_schema("postgres", "customers", schema: "reporting")
{:ok, schema} = Lotus.get_table_schema(MyApp.DataRepo, "products", search_path: "analytics, public")
"""
def get_table_schema(repo_or_name, table_name, opts \\ []),
do: Schema.get_table_schema(repo_or_name, table_name, opts)
@doc """
Gets statistics for a specific table.
## Examples
{:ok, stats} = Lotus.get_table_stats("primary", "users")
{:ok, stats} = Lotus.get_table_stats("postgres", "customers", schema: "reporting")
# Returns %{row_count: 1234}
"""
def get_table_stats(repo_or_name, table_name, opts \\ []),
do: Schema.get_table_stats(repo_or_name, table_name, opts)
@doc """
Lists all relations (tables with schema information) in a data repository.
## Examples
{:ok, relations} = Lotus.list_relations("postgres", search_path: "reporting, public")
# Returns [{"reporting", "customers"}, {"public", "users"}, ...]
"""
def list_relations(repo_or_name, opts \\ []), do: Schema.list_relations(repo_or_name, opts)
defp cache_mode(nil) do
case Config.cache_adapter() do
{:ok, _adapter} -> :use
:error -> :off
end
end
defp cache_mode(:bypass), do: :bypass
defp cache_mode(:refresh), do: :refresh
defp cache_mode(opts) when is_list(opts) do
cond do
:bypass in opts -> :bypass
:refresh in opts -> :refresh
true -> :use
end
end
defp choose_ttl(cache_opts, default_profile) do
cond do
# caller explicitly passed a ttl override
is_list(cache_opts) and Keyword.has_key?(cache_opts, :ttl_ms) ->
Keyword.fetch!(cache_opts, :ttl_ms)
true ->
# resolve profile in this order:
# 1. caller passed `:profile`
# 2. fallback from caller (e.g. :results / :schema)
# 3. globally configured default profile
prof =
cond do
is_list(cache_opts) and Keyword.has_key?(cache_opts, :profile) ->
Keyword.fetch!(cache_opts, :profile)
not is_nil(default_profile) ->
default_profile
true ->
Config.default_cache_profile()
end
Config.cache_profile_settings(prof)[:ttl_ms] ||
(Config.cache_config() && Config.cache_config()[:default_ttl_ms]) ||
:timer.seconds(60)
end
end
defp exec_with_cache(cache_opts, ttl_default_profile, key, tags, fun) do
case cache_mode(cache_opts) do
:off ->
fun.()
:bypass ->
fun.()
:refresh ->
case fun.() do
{:ok, val} ->
ttl = choose_ttl(cache_opts, ttl_default_profile)
cache_options = build_cache_options(cache_opts, tags)
:ok = Lotus.Cache.put(key, val, ttl, cache_options)
{:ok, val}
error ->
error
end
:use ->
ttl = choose_ttl(cache_opts, ttl_default_profile)
try do
cache_options = build_cache_options(cache_opts, tags)
case Lotus.Cache.get_or_store(
key,
ttl,
fn ->
case fun.() do
{:ok, val} -> val
{:error, e} -> throw({:lotus_cache_error, e})
end
end,
cache_options
) do
{:ok, val, _} -> {:ok, val}
{:error, _} -> fun.()
end
catch
{:lotus_cache_error, e} -> {:error, e}
end
end
end
defp build_cache_options(cache_opts, tags) do
base_options = [tags: tags]
if is_list(cache_opts) do
cache_options =
cache_opts
|> Enum.filter(fn
{_key, _value} -> true
_atom -> false
end)
|> Keyword.take([:max_bytes, :compress])
Keyword.merge(cache_options, base_options)
else
base_options
end
end
defp result_key(sql, bound_vars_map, repo_name, search_path) do
Lotus.Cache.Key.result(sql, bound_vars_map,
data_repo: repo_name,
search_path: search_path,
lotus_version: Lotus.version()
)
end
defp resolve_window_limit(window_opts) do
max_limit = Config.default_page_size() || @default_page_size
case Keyword.get(window_opts, :limit) do
nil ->
max_limit
limit when not is_integer(limit) or limit <= 0 ->
max_limit
limit ->
min(limit, max_limit)
end
end
defp maybe_apply_window(sql, params, _repo_or_name, _search_path, nil),
do: {sql, params, nil, nil}
defp maybe_apply_window(sql, params, repo_or_name, search_path, window_opts)
when is_list(window_opts) do
base_sql = trim_trailing_semicolon(sql)
limit = resolve_window_limit(window_opts)
offset = Keyword.get(window_opts, :offset, 0)
count_mode = Keyword.get(window_opts, :count, :none)
{limit_ph, offset_ph} =
Lotus.Source.limit_offset_placeholders(
repo_or_name,
length(params) + 1,
length(params) + 2
)
paged_sql =
"SELECT * FROM (" <>
base_sql <> ") AS lotus_sub LIMIT " <> limit_ph <> " OFFSET " <> offset_ph
paged_params = params ++ [limit, offset]
window_meta =
case count_mode do
:exact ->
%{
window: %{limit: limit, offset: offset},
total_count: :pending,
total_mode: :exact,
count_sql: "SELECT COUNT(*) FROM (" <> base_sql <> ") AS lotus_sub",
count_params: params,
repo_or_name: repo_or_name,
search_path: search_path
}
_ ->
%{window: %{limit: limit, offset: offset}, total_count: nil, total_mode: :none}
end
# Include window in cache key bound variables
cache_bound = %{
__params__: params,
__window__: %{limit: limit, offset: offset, count: count_mode}
}
{paged_sql, paged_params, window_meta, cache_bound}
end
defp merge_window_meta(%Result{} = res, nil), do: res
defp merge_window_meta(%Result{} = res, %{total_mode: :none, window: win} = _meta) do
updated_meta = Map.merge(res.meta, %{window: win})
%Result{res | num_rows: length(res.rows), meta: updated_meta}
end
defp merge_window_meta(%Result{} = res, %{total_mode: :exact} = meta) do
win = Map.fetch!(meta, :window)
# Try to compute exact count synchronously. If it fails, fall back with no total.
total_count =
case do_count(meta) do
{:ok, n} when is_integer(n) and n >= 0 -> n
_ -> nil
end
updated_meta =
Map.merge(res.meta, %{window: win, total_count: total_count, total_mode: :exact})
%Result{
res
| num_rows: length(res.rows),
meta: updated_meta
}
end
defp do_count(
%{count_sql: count_sql, count_params: count_params, repo_or_name: repo_or_name} = meta
) do
{repo_mod, _repo_name} = Sources.resolve!(repo_or_name, nil)
runner_opts = []
runner_opts =
case Map.get(meta, :search_path) do
sp when is_binary(sp) and byte_size(sp) > 0 -> Keyword.put(runner_opts, :search_path, sp)
_ -> runner_opts
end
case Runner.run_sql(repo_mod, count_sql, count_params, runner_opts) do
{:ok, %Result{rows: [[n]]}} when is_integer(n) ->
{:ok, n}
{:ok, %Result{rows: [[n]]}} when is_binary(n) ->
case Integer.parse(n) do
{v, _} -> {:ok, v}
_ -> {:error, :invalid_count}
end
{:ok, %Result{rows: _}} ->
{:error, :invalid_count}
other ->
other
end
end
defp trim_trailing_semicolon(sql) do
s = String.trim(sql)
if String.ends_with?(s, ";") do
s
|> String.trim_trailing()
|> String.trim_trailing(";")
|> String.trim_trailing()
else
s
end
end
end