[ghc-steering-committee] Proposal #412: Explicit Splice Imports; rec: accept

[Vladislav Zavialov (int-index)]

Dear Committee,

Proposal #412 "Explicit Splice Imports” by Matthew Pickering has been submitted for our consideration.

Read it here: https://github.com/mpickering/ghc-proposals/blob/splice-imports/proposals/0000-splice-imports.rst
Discussion here: https://github.com/ghc-proposals/ghc-proposals/pull/412

The goals of this proposal are to (1) improve parallel compilation, (2) speed up -fno-code and IDEs that rely on it, (3) reduce binary sizes or at least reduce linking overhead, (4) create phase separation useful for cross-compilation.

All of the above are achieved by introducing a new form of import called a splice-import. Compare:

  import M
  import splice H

Under this proposal, names imported from M can *not* be used in top-level Template Haskell splices, whereas names imported from H can be used *only* in Template Haskell splices.

At this point I will note that I do not particularly like the proposed syntax, which looks as if we’re importing a splice rather than importing names for use in splices. In the “Alternatives” section, you will find several other options such as “import {-# SPLICE #-} M”, “import for splice H”, “$(import H)”, and so on. That said, let us not get bogged down in the discussion of syntax before we agree on the semantics. If we choose to accept the proposal, I will organize a vote on the syntax.

Regarding the semantics, I believe the general idea behind the proposal is good, and I recommend acceptance. It provides a way for the programmer to specify the intended usage of an import (whether it is for normal programming or metaprogramming), and GHC can use this information to build the program more effectively (more parallelism, less linking). In ways that I do not entirely understand, it is allegedly good for future work on cross-compilation.

Let me know what you think.

- Vlad