How faces and edges are named
A feature that picks a face or an edge (a sketch on a face, a pad up to a face, a fillet's edges, a joint's faces) keeps the face's or edge's name beside the point where it was picked. When the part rebuilds, the face or edge is found by its name first, and by the point only when the name finds nothing. This is what keeps a fillet on its edge when a dimension earlier in the history changes, or when a feature goes in before it.
This page is the rules those names follow, so a reference that lands
on the wrong face can be reported precisely. Scripts read names through
doc.faces and doc.edges (as strings: a name is
a 64-bit number, more than a script's numbers hold exactly).
What a name is
A name is a 64-bit number made from what it names, the same on every
machine and in every version, because documents keep it
(kernel_api::naming). Zero is no name. A name is made from
a parent name and a part (child(parent, part)): a feature's
name and the sketch element a face was swept from, or a feature's name
and the word end.
The rules
- A feature names the faces it makes from what made
them. Each wall a sketch's segment sweeps is named after that
segment (the sketch element it was drawn as,
ProfileWire::names), so the walls keep their names while the sketch's dimensions change. - A feature's ends are named by where they stand
along the sketch's normal:
starton the sketch's plane,endpast it,backbehind it (a feature built both ways). - Any other face a feature makes is named by its kind of surface and the way it faces, ties among alike faces broken by where each lies. A box keeps its six names when it grows.
- A face an operation leaves as it was keeps its names, however the kernel stores it afterwards: recognised by its kind of surface, its edges and vertices and where they lie.
- A face an operation splits, trims or merges takes the names of every face it came from. The kernel's own record of the operation says which; where it says nothing, a piece is named after the face it lies on, and a merged face after every face it covers. One face can so carry several names, and a split leaves several faces with one name.
- A face made from an edge (a fillet's round, a chamfer's band) is named after that edge's two faces.
- An edge is named by the two faces it runs between, by their least names.
- Finding a face by name: the face bearing that name; among several (a face a later feature split), the one nearest the point the reference was picked at. Finding an edge: an edge between faces of the two names; among several, the one passing nearest the point.
- A pick keeps one name of a face, its least: enough to find it, and the same whichever way the face was reached.
Where each rule is held
| Rule | Code | Test (crates/kernel_ogeom/tests/kernel/naming.rs) |
|---|---|---|
| 1, 2, 3 | naming::tool_names,
NameMap::name_the_rest |
every_face_is_named_and_the_names_hold_across_builds |
| 4, 5 | NameMap::carry |
an_up_to_face_follows_its_face_by_name,
a_thickness_opens_its_face_by_name |
| 6 | NameMap::carry (an edge's faces) |
a_fillet_s_round_is_named_after_its_edge |
| 7, 9 | naming::edge_faces, NameMap::primary |
a_fillet_s_round_is_named_after_its_edge |
| 8 | naming::find_face, naming::find_edge |
a_face_probe_finds_its_face_by_name |
Reporting a reference that moved
Say which feature's reference moved, what it was picked on, and what
changed before it moved (a dimension, a feature added before it). The
names of the faces involved, from doc.faces before and
after, show whether the face kept its name (a lookup problem, rule 8) or
lost it (a naming problem, rules 1 to 7).