defmodule SearchAsh.MixProject do use Mix.Project @version "0.4.0" @source_url "https://github.com/nseaSeb/search_ash" def project do [ app: :search_ash, version: @version, elixir: "~> 1.19", start_permanent: Mix.env() == :prod, elixirc_paths: elixirc_paths(Mix.env()), description: "Ash extension for multilingual full-text search: a `search do … end` DSL that " <> "auto-generates the tsvector index, keeps a stemmed column in sync, and exposes " <> "a tenant-aware, ranked `search` action. Built on search_core.", package: package(), docs: docs(), test_coverage: test_coverage(), source_url: @source_url, deps: deps() ] end # Run "mix help compile.app" to learn about applications. def application do [ extra_applications: [:logger] ] end defp elixirc_paths(:test), do: ["lib", "test/support"] defp elixirc_paths(_), do: ["lib"] # Run "mix help deps" to learn about dependencies. defp deps do [ search_core_dep(), {:ash, "~> 3.29"}, {:ash_postgres, "~> 2.10"}, {:ex_doc, "~> 0.34", only: :dev, runtime: false}, {:sourceror, "~> 1.0", only: [:dev, :test], runtime: false}, # Ash policies need a SAT solver; simple_sat is pure Elixir, fitting for a NIF-free stack. {:simple_sat, "~> 0.1", only: :test} ] end # Inside the monorepo the sibling source is what we want to build against — a change to # search_core should be visible here before it is published. Set SEARCH_ASH_LOCAL_CORE=1 # for that; the published package always resolves through Hex. defp search_core_dep do if System.get_env("SEARCH_ASH_LOCAL_CORE") in ~w(1 true) do {:search_core, path: "../search_core", override: true} else {:search_core, "~> 0.3"} end end defp package do [ licenses: ["MIT"], links: %{"GitHub" => @source_url}, files: [ "lib", "mix.exs", "README.md", "LICENSE", ".formatter.exs", "CHANGELOG.md", "documentation" ] ] end # Cover only traces modules loaded at *test* time, which shapes what this number can # honestly say: # # * Ignored below: derived `Inspect` impls, the test fixtures themselves, and the # structs Spark generates for DSL options and entity targets. None of it is code # this project wrote or could test. # * NOT ignored, though they report 0%: the DSL transformers. They run at *compile* # time, when every fixture resource is built — so they are exercised by the whole # suite, just not measurably. (The per-resource ones score higher only because # `verify_language_test.exs` compiles resources inside test bodies.) Hiding them # would overstate the number; leaving them in understates it. The honest reading is # that runtime modules sit in the 90s. defp test_coverage do [ # NB: the threshold lives under `:summary`, not at the top level — Mix reads it with # `get_threshold(summary_opts)`, so a bare `threshold:` here is silently ignored and # you get the 90% default. summary: [threshold: 70], ignore_modules: [ ~r/^Inspect\./, ~r/^SearchAsh\.Test\./, ~r/\.Options$/, ~r/^SearchAsh\.Source\.Searchable\./ ] ] end defp docs do [ main: "SearchAsh", source_ref: "v#{@version}", source_url: @source_url, extras: [ "README.md", "documentation/global-search.md", "documentation/architecture.md", "documentation/upgrading-0.4.md", "documentation/roadmap.md" ], # Render ```mermaid fences on HexDocs (GitHub renders them natively). before_closing_body_tag: fn :html -> """ """ _ -> "" end ] end end