[C API] Metadata schema field on tables

[benjeffery]

To split #491 up a bit I'm focusing on getting the C API right first.
Proof-of-concept for the python is done in #491 meaning I'm pretty sure this is the (per-table) changes we want. Only individuals is modified for now. Once everyone is happy I'll do the other tables.
I intend to do top-level (per tree sequence) metadata in another PR.

• Agree on pre-table changes
• Individuals
• Nodes
• Edges (including adding metadata column (done in Add metadata column to edges #496))
• Sites
• Mutations (including adding metadata column (done in Add metadata column to the migrations table #505)
• Migrations
• Populations

[benjeffery]

@jeromekelleher You mentioned text dump round tripping - is that in python only? Couldn't see it in the C.

[jeromekelleher]

@jeromekelleher You mentioned text dump round tripping - is that in python only? Couldn't see it in the C.

Yep. Text support in C is strictly just for testing. Life is too short to write production quality text parsing in C!

[jeromekelleher]

LGTM @benjeffery. Minor points above.

[codecov]

Codecov Report

Merging #493 into master will decrease coverage by 0.11%.
The diff coverage is 73.55%.

@@            Coverage Diff             @@
##           master     #493      +/-   ##
==========================================
- Coverage   86.96%   86.84%   -0.12%     
==========================================
  Files          21       21              
  Lines       15748    15944     +196     
  Branches     3066     3095      +29     
==========================================
+ Hits        13695    13847     +152     
- Misses       1034     1050      +16     
- Partials     1019     1047      +28     

Flag	Coverage Δ
#c_tests	87.77% <73.55%> (-0.18%)	⬇️
#python_c_tests	90.52% <ø> (ø)
#python_tests	99.18% <ø> (ø)

Impacted Files	Coverage Δ
c/tskit/tables.c	77.73% <73.55%> (-0.01%)	⬇️

---

Continue to review full report at Codecov.

Legend - Click here to learn more
Δ = absolute <relative> (impact), ø = not affected, ? = missing data
Powered by Codecov. Last update f53ff61...363b49d. Read the comment docs.

[jeromekelleher]

LGTM, ready to copy-paste across many tables I think!

[benjeffery]

@jeromekelleher I've given this another pass, after what we talked about with edge metadata. I've made metadata_schema an optional load and memcpy'd it out of kastore.
This still has metadata_schema == NULL for empty schemas. I think this makes sense for this, where it didn't for columns.

[jeromekelleher]

LGTM @benjeffery, I'd say go for it. The NULL vs non-NULL thing is a bit philosophical when metadata_length == 0, so I think setting it to NULL on 0 is the right thing to do.

[benjeffery]

Ok, will copy across to other tables now.

[jeromekelleher]

LGTM @benjeffery, just spotted a minor typo in the docs. Shall we merge this much and follow up with the Python tests?

[benjeffery]

@jeromekelleher Squashed rebased and ready to merge (pending CI!). Yes, lets merge here as this is a natural break point as nothing is broken, just added.

[benjeffery]

C lint has failed as azure is on its knees. I don't have permission to re-run it.

[jeromekelleher]

C lint has failed as azure is on its knees. I don't have permission to re-run it.

I've been getting these errors a bit elsewhere too - tried rerunning them but they kept failing.

[benjeffery]

I'll see if we can remove the dependency