💾 Archived View for thebird.nl › blog › 2023 › mgamma.gmi captured on 2023-05-24 at 17:34:33. Gemini links have been rewritten to link to archived content
⬅️ Previous capture (2023-03-20)
-=-=-=-=-=-=-
It is February 2023 and, after some detours in pangenomics and vcflib, I am back on linear mixed models (LMMs). I have decided the time is right to pursue a new tool named `mgamma`. Mostly because mgamma sounds nice and may be about right for people in genetical genomics ;).
Mgamma will be written in zig with a dash of guile, mostly because these are minimalistic languages that supplement each other. I am taking a cue from Joe Armstrong who said that part of the success of Erlang was its few dependencies, even so I am settling for Zig and guile. Development will be hosted on our own git instance.
The first step for mgamma is to read and write lmdb files. Thanks to work by Arun Isaac we can store versioned and deduplicated vectors in an lmdb file. I created a build envirnonment that includes lmdb as a shared lib. It can simply be passed in as `-llmdb`.
Next, I tried compiling
https://github.com/lithdew/lmdb-zig/
but it is not even happy with zig version 0.9. One particular error
contrib/lmdb-zig/lmdb.zig:749:23: error: expected type 'builtin.CallModifier', found '@TypeOf(.{})'
const rc = @call(.{}, function, args);
is probably easy to fix, but I have to see if I want to carry the lmdb-zig code along as it is a simple C library. Zig has a translation function:
zig translate-c $GUIX_ENVIRONMENT/include/lmdb.h > lmdb.zig
and it compiles. Only problem is that I am getting now
mgamma.zig:22:9: error: variable of type 'type' must be const or comptime
var env = [*c]?*lmdb.MDB_env;
^~~
I'll have to look at the C code and maybe roll my own opaque pointer. The lmdb-zig package uses
try call(c.mdb_env_create, .{&inner});
(note the ampersand) where
inner: ?*c.MDB_env
Meanwhile lmdb simply says
int mdb_env_create(MDB_env **env)
How hard can it be? The lmdb docs are at
http://www.lmdb.tech/doc/group__mdb.html
This led to a blog on zig pointers
This CL code builds an lmdb genotype file
https://github.com/genenetwork/cl-gn/blob/main/genodb.lisp
it exports
:genotype-db-current-matrix :genotype-db-matrix :genotype-db-matrix-ref :genotype-db-matrix-row-ref :genotype-db-matrix-column-ref
In the next phase we are going to compute the kinship matrix: