A lookup from a dimension into itself, so representative snapshots are sayable

[FBumann]

Prompt: "And yes, open an issue about the self-mapping"

Note

The following content was generated by AI. Filed from the discussion under #275, where splitting lookups: into groupers and connectors was explored and rejected; this is the one capability the word "connector" reaches for that the language refuses.

What can be improved?

A lookup must map into a dimension other than the one it is over. Today:

lookups:
  rep_of: { over: snapshot, into: snapshot }

is refused at load with Lookup 'rep_of' maps 'snapshot' into itself. A lookup maps into a different dimension.

That is the representative-period map (tsam, Kotzur): every snapshot names the representative snapshot that stands for it, and #6 listed it as the second kind of "connector". With the refusal lifted it needs no new construct:

call	reads
at(x, by=rep_of)	the representative's value at every snapshot it stands for
sum(x, by=rep_of)	x summed onto each representative from the snapshots it stands for
shift(x, over=snapshot, offset=1, by=rep_of)	the walk inside each representative's group
position(snapshot, by=rep_of) == 0	the first snapshot of each group
where: rep_of	the snapshots a representative stands for

The dim algebra holds with one added sentence: where over and into coincide, the operator consumes and produces the same dimension, so the frame is unchanged. sum(by=) and at(by=) stay adjoints. The bind-time checks are verbatim: values are labels of snapshot, one row per snapshot.

One open question. The map's table today carries two columns, one named after over and one after into. For a self-map both are snapshot, so the value column needs another name. The candidates are the lookup's own name (rep_of), or a fixed convention such as into; the reference's "the map is supplied under the lookup's own name" section would say which.

per: (#161) composes with this unchanged: a representative that varies by period is {over: snapshot, into: snapshot, per: [period]}.

What lifting it touches

• model.py: the into == over refusal.
• dimensions.py: _sum_dims and _at_dims, whose "would need the dim twice" checks read over and into as distinct.
• the typesetter legend, which prints rep_of: 𝒯 → 𝒯 without change.
• a where comparing two lookups over and into snapshot already works.

Version

0.0.0-alpha.80, with #422 applied.

[FBumann]

Prompt: "#437 was merged / Close all related issues and stale PRs"

Note

The following content was generated by AI.

Shipped by #437, merged as 3284926 and released in v0.0.0-alpha.89. Its body names this issue as covered.

A self-map is a relation with two columns over one dimension, told apart by roles:

relations:
  rep: { columns: { snapshot: snapshot, rep: snapshot }, key: snapshot }

The mapping form of columns: exists for exactly this and for a line's two ends — a list names columns after their dimensions, and a mapping gives them names where two columns share one. That is why the declaration says columns: and not dims:: a frame is a product of distinct dimensions, and a relation's columns are not.

One limit is worth carrying forward rather than losing here: key: cannot narrow the bind check to one-to-one. The check stays many-to-one, so a representative-snapshot table is held to each snapshot appearing once and not to each representative being used once. #437 records several candidate keys as deferred.

#436 answered this on top of #433 and is closed with that proposal.

---

Generated by Claude Code