Skip to content

Latest commit

 

History

History
109 lines (70 loc) · 4.52 KB

DOC.md

File metadata and controls

109 lines (70 loc) · 4.52 KB
Raw

Documentation for MetaCoq users and developers

Branches and compatibility

tl;dr You should do your PRs against coq-8.18.

Coq's kernel API is not stable yet, and changes there are reflected in MetaCoq's reified structures, so we do not ensure any compatibility from version to version. There is one branch for each Coq version.

The main branch or current branch is the one which appers when you go on https://github.com/MetaCoq/metacoq. Currently (unless you are reading the README of an outdated branch), it is the coq-8.16. You should use it both for usage of MetaCoq and development of MetaCoq.

The main branch is following Coq's master branch and gets regular updates from the the main development branch which follows the latest stable release of Coq.

The branches coq-8.17 and coq-8.18 are being kept in sync. The branches coq-8.6 to coq-8.15 are frozen.

Program and Equations

MetaCoq relies on Program and Equations plugins, however try to avoid Program as it inserts some JMeq and UIP axioms silently, whereas we try to keep the development axiom-free. You can use Equations to do some dependent induction (dependent induction, dependent destruction, depelim). You may need to add:

Require Import Equations.Prop.DepElim.

ident vs. qualid. vs kername

MetaCoq uses three types convertible to string which have a different intended meaning:

  • ident is the type of identifiers, they should not contains any dot. E.g. nat

  • qualid is the type of partially qualified names. E.g. Datatypes.nat

  • kername is a structured type of fully qualified names. E.g. Coq.Init.Datatypes.nat

Quoting always produce fully qualified names. On the converse, unquoting allow to have only partially qualified names and rely on Coq to resolve them. The commands of the TemplateMonad also allow partially qualified names.

Hint databases

The development uses three main hint databases:

  • The "len" databases which gathers all relevant length lemmas (mainly list length lemmas relevant to the operations). This database is large (> 100 lemmas) for a given theory (PCUIC or Template-Coq) and it is advisable to not mix together both databases, as autorewrite would become very slow. BEWARE: If working in the PCUIC theory, do not require anything besides the BasicAst and utils modules from the Template-Coq module.
  • The "pcuic" rewrite and auto database gathers lemmas helping solving side-conditions of typing judgements.

Options

Set / Unset MetaCoq Strict Unquote Universe Mode. When this mode is on, the unquoting of a universe level fails if this level does not exists. Otherwise the level is added to the current context. It is on by default.

There is also an "opaque" universe fresh_universe which is unquoted to a fresh level when MetaCoq Strict Unquote Universe Mode is off.

Contribution guidelines

  • Please use as many bullets as possible. You even can be forced to do so with Set Default Goal Selector "!".

  • Please use as few as possible generated names and name hypothesis in intros and destruct. It is more difficult for induction and above all for inversion.

Dependency graph between files

Generated on 2022/07/01, sources there.

Dependency graph

MetaDocumentation (documentation about documentation)

The file README.md in https://github.com/MetaCoq/metacoq.github.io is supposed to be synchronized with README.md in https://github.com/MetaCoq/metacoq/.

That's why we can't use relative links and have to use absolute ones. E.g. INSTALL.md and not INSTALL.md.

Thus, when switching to a new default branch, we have to search and replace the old branch with the new one.