08-generic-builders.xml

<chapter xmlns="http://docbook.org/ns/docbook"
      xmlns:xlink="http://www.w3.org/1999/xlink"
      xmlns:xi="http://www.w3.org/2001/XInclude"
      version="5.0"
      xml:id="generic-builders">

  <title>Generic Builders</title>

  <para>
    Welcome to the 8th Nix pill. In the previous
    <link linkend="working-derivation">7th pill</link> we successfully built a
    derivation. We wrote a builder script that compiled a C file and installed
    the binary under the nix store.
  </para>

  <para>
    In this post, we will generalize the builder script, write a Nix expression
    for <link
    xlink:href="http://www.gnu.org/software/hello/">GNU hello world</link>
    and create a wrapper around the derivation built-in function.
  </para>

  <section>
    <title>Packaging GNU hello world</title>

    <para>
      In the previous pill we packaged a simple .c file, which was being
      compiled with a raw gcc call. That's not a good example of a project. Many
      use autotools, and since we're going to generalize our builder, it would
      be better do it with the most used build system.
    </para>

    <para>
      <link xlink:href="http://www.gnu.org/software/hello/">GNU hello world</link>,
      despite its name, is a simple yet complete project which uses autotools.
      Fetch the latest tarball here:
      <link xlink:href="http://ftp.gnu.org/gnu/hello/hello-2.10.tar.gz">http://ftp.gnu.org/gnu/hello/hello-2.10.tar.gz</link>.
    </para>

    <para>
      Let's create a builder script for GNU hello world, hello_builder.sh:
    </para>

    <screen><xi:include href="./08/hello-builder.txt" parse="text" /></screen>

    <para>
      And the derivation hello.nix:
    </para>

    <screen><xi:include href="./08/hello-nix.txt" parse="text" /></screen>
    <note><title>Nix on darwin</title>
    <para>Darwin (i.e. macOS) builds typically use <literal>clang</literal> rather than <literal>gcc</literal> for a C compiler.
    We can adapt this early example for darwin by using this modified version of <filename>hello.nix</filename>:
    <screen><xi:include href="./08/hello-nix-darwin.txt" parse="text" /></screen>
    Later, we will show how Nix can automatically handle these differences.
    For now, please be just aware that changes similar to the above may be needed in what follows.
    </para>
    </note>

    <para>
      Now build it with <command>nix-build hello.nix</command> and you can
      launch <filename>result/bin/hello</filename>. Nothing easier, but do we
      have to create a builder.sh for each package? Do we always have to pass
      the dependencies to the <literal>derivation</literal> function?
    </para>

    <para>
      Please note the <command>--prefix=$out</command> we were talking about in
      the <link linkend="working-derivation">previous pill</link>.
    </para>
  </section>

  <section>
    <title>A generic builder</title>

    <para>
      Let's create a generic <filename>builder.sh</filename> for autotools
      projects:
    </para>

    <screen><xi:include href="./08/generic-builder.txt" parse="text" /></screen>

    <para>
      What do we do here?
    </para>

    <orderedlist>
      <listitem>
        <para>
          Exit the build on any error with <command>set -e</command>.
        </para>
      </listitem>
      <listitem>
        <para>
          First <command>unset PATH</command>, because it's initially set to a
          non-existant path.
        </para>
      </listitem>
      <listitem>
        <para>
          We'll see this below in detail, however for each path in
          <code>$buildInputs</code>, we append <code>bin</code> to
          <code>PATH</code>.
        </para>
      </listitem>
      <listitem>
        <para>
          Unpack the source.
        </para>
      </listitem>
      <listitem>
        <para>
          Find a directory where the source has been unpacked and
          <command>cd</command> into it.
        </para>
      </listitem>
      <listitem>
        <para>
          Once we're set up, compile and install.
        </para>
      </listitem>
    </orderedlist>

    <para>
      As you can see, there's no reference to "hello" in the builder anymore.
      It still does several assumptions, but it's certainly more generic.
    </para>

    <para>
      Now let's rewrite <filename>hello.nix</filename>:
    </para>

    <screen><xi:include href="./08/hello-nix-rev-1.txt" parse="text" /></screen>

    <para>
      All clear, except that buildInputs. However it's easier than any black
      magic you are thinking in this moment.
    </para>

    <para>
      Nix is able to convert a list to a string. It first converts the elements
      to strings, and then concatenates them separated by a space:
    </para>

    <screen><xi:include href="./08/to-string.txt" parse="text" /></screen>

    <para>
      Recall that derivations can be converted to a string, hence:
    </para>

    <screen><xi:include href="./08/to-string-nixpkgs.txt" parse="text" /></screen>

    <para>
      Simple! The buildInputs variable is a string with out paths separated by
      space, perfect for bash usage in a for loop.
    </para>
  </section>

  <section>
    <title>A more convenient derivation function</title>

    <para>
      We managed to write a builder that can be used for multiple autotools
      projects. But in the hello.nix expression we are specifying tools that
      are common to more projects; we don't want to pass them everytime.
    </para>

    <para>
      A natural approach would be to create a function that accepts an
      attribute set, similar to the one used by the derivation function, and
      merge it with another attribute set containing values common to many
      projects.
    </para>

    <para>
      Create <filename>autotools.nix</filename>:
    </para>

    <screen><xi:include href="./08/autotools-nix.txt" parse="text" /></screen>

    <para>
      Ok now we have to remember a little about
      <link linkend="functions-and-imports">Nix functions</link>. The whole nix
      expression of this <filename>autotools.nix</filename> file will evaluate
      to a function. This function accepts a parameter <code>pkgs</code>, then
      returns a function which accepts a parameter <code>attrs</code>.
    </para>

    <para>
      The body of the function is simple, yet at first sight it might be hard
      to grasp:
    </para>

    <orderedlist>
      <listitem>
        <para>
          First drop in the scope the magic <code>pkgs</code> attribute set.
        </para>
      </listitem>
      <listitem>
        <para>
          Within a let expression we define a helper variable,
          <code>defaultAttrs</code>, which serves as a set of common attributes
          used in derivations.
        </para>
      </listitem>
      <listitem>
        <para>
          Finally we create the derivation with that strange expression,
          (<code>defaultAttrs // attrs</code>).
        </para>
      </listitem>
    </orderedlist>

    <para>
      The
      <link xlink:href="https://nixos.org/manual/nix/stable/#sec-language-operators">// operator</link>
      is an operator between two sets. The result is the union of the two sets.
      In case of conflicts between attribute names, the value on the right set
      is preferred.
    </para>

    <para>
      So we use <code>defaultAttrs</code> as base set, and add (or override) the
      attributes from <code>attrs</code>.
    </para>

    <para>
      A couple of examples ought to be enough to clear out the behavior of the
      operator:
    </para>

    <screen><xi:include href="./08/set-union.txt" parse="text" /></screen>

    <para>
      <emphasis role="bold">Exercise:</emphasis>
      Complete the new <filename>builder.sh</filename> by adding
      <code>$baseInputs</code> in the <code>for</code> loop together with
      <code>$buildInputs</code>. As you noticed, we passed that new variable in
      the derivation. Instead of merging buildInputs with the base ones, we
      prefer to preserve buildInputs as seen by the caller, so we keep them
      separated. Just a matter of choice.
    </para>

    <para>
      Then we rewrite <filename>hello.nix</filename> as follows:
    </para>

    <screen><xi:include href="./08/hello-nix-rev-2.txt" parse="text" /></screen>

    <para>
      Finally! We got a very simple description of a package! Below are a
      couple of remarks that you may find useful as you're continuing to
      understand the nix language:
    </para>

    <itemizedlist>
      <listitem>
        <para>
          We assigned to pkgs the import that we did in the previous expressions
          in the "with". Don't be afraid, it's that straightforward.
        </para>
      </listitem>
      <listitem>
        <para>
          The mkDerivation variable is a nice example of partial application,
          look at it as (<code>import ./autotools.nix</code>) <code>pkgs</code>.
          First we import the expression, then we apply the <code>pkgs</code>
          parameter. That will give us a function that accepts the attribute
          set <code>attrs</code>.
        </para>
      </listitem>
      <listitem>
        <para>
          We create the derivation specifying only name and src. If the project
          eventually needed other dependencies to be in PATH, then we would
          simply add those to buildInputs (not specified in hello.nix because
          empty).
        </para>
      </listitem>
    </itemizedlist>

    <para>
      Note we didn't use any other library. Special C flags may be needed to
      find include files of other libraries at compile time, and ld flags at
      link time.
    </para>
  </section>

  <section>
    <title>Conclusion</title>

    <para>
      Nix gives us the bare metal tools for creating derivations, setting up a
      build environment and storing the result in the nix store.
    </para>

    <para>
      Out of this pill we managed to create a generic builder for autotools
      projects, and a function <code>mkDerivation</code> that composes by default
      the common components used in autotools projects instead of repeating them
      in all the packages we would write.
    </para>

    <para>
      We are familiarizing ourselves with the way a Nix system grows up: it's
      about creating and composing derivations with the Nix language.
    </para>

    <para>
      <emphasis role="underline">Analogy</emphasis>: in C you create objects
      in the heap, and then you compose them inside new objects. Pointers are
      used to refer to other objects.
    </para>

    <para>
      In Nix you create derivations stored in the nix store, and then you
      compose them by creating new derivations. Store paths are used to refer
      to other derivations.
    </para>
  </section>

  <section>
    <title>Next pill</title>

    <para>
      ...we will talk a little about runtime dependencies. Is the GNU hello
      world package self-contained? What are its runtime dependencies? We only
      specified build dependencies by means of using other derivations in the
      "hello" derivation.
    </para>
  </section>
</chapter>