Is your feature request related to a problem?
There's no way to size an element relative to its parent as a constraint rather than a fixed size. We can say "exactly 50% of the parent" with percent(0.5), but not "at least 50%" or "at most 50%" while still letting content drive the actual size. A sidebar that should be ≥50% tall but grow with its content, or a panel capped at half the viewport, can't be expressed: percent(0.5) pins to exactly 6 rows in a 12-row parent regardless of content, and fit()/grow() only take absolute cell counts in their min/max slots. (Passing a percent into a min/max slot today corrupts layout—it's coerced into a garbage float.)
Describe the solution you'd like
A way to express parent-relative floors and caps, not just fixed sizes, so an element can be bounded by a fraction of its parent while content still determines the size within those bounds. The exact API is open; a few directions worth weighing:
- Percent in the existing min/max slots —
fit(percent(0.5)) / fit(0, percent(0.5)). Smallest surface, but overloads the min/max args.
- Dedicated clamp helpers —
clamp(min, max) or minmax(min, max) accepting cells or percent(...), composing with fit()/grow(). More explicit about intent.
min() / max() constraint helpers — CSS-like, e.g. max(fit(), percent(0.5)). Most general, largest surface.
Whatever the shape: numeric bounds must keep exact absolute-cell semantics (additive only), and a percent bound should clamp content-aware—a floor still lets content grow past it, a cap still lets content shrink below it.
Describe alternatives you've considered
Measure the parent first (info.get(parent).bounds), compute the cell count, and feed an absolute number back on a second pass—but that needs two renders, goes stale on resize, and can't express floor-plus-content-grows in one pass. The whole-axis percent() is the existing primitive that falls short here.
Additional context
Failing test case on nm/repro/percent-sizing (test · diff). Whichever shape wins, the clamp lives deepest in the layout engine: Clay_SizingMinMax (clay/clay.h:316-328) is absolute float cells with no parent-relative bound, so it needs a percent-aware min/max on the wire and in the engine.
Is your feature request related to a problem?
There's no way to size an element relative to its parent as a constraint rather than a fixed size. We can say "exactly 50% of the parent" with
percent(0.5), but not "at least 50%" or "at most 50%" while still letting content drive the actual size. A sidebar that should be ≥50% tall but grow with its content, or a panel capped at half the viewport, can't be expressed:percent(0.5)pins to exactly 6 rows in a 12-row parent regardless of content, andfit()/grow()only take absolute cell counts in their min/max slots. (Passing a percent into a min/max slot today corrupts layout—it's coerced into a garbage float.)Describe the solution you'd like
A way to express parent-relative floors and caps, not just fixed sizes, so an element can be bounded by a fraction of its parent while content still determines the size within those bounds. The exact API is open; a few directions worth weighing:
fit(percent(0.5))/fit(0, percent(0.5)). Smallest surface, but overloads the min/max args.clamp(min, max)orminmax(min, max)accepting cells orpercent(...), composing withfit()/grow(). More explicit about intent.min()/max()constraint helpers — CSS-like, e.g.max(fit(), percent(0.5)). Most general, largest surface.Whatever the shape: numeric bounds must keep exact absolute-cell semantics (additive only), and a percent bound should clamp content-aware—a floor still lets content grow past it, a cap still lets content shrink below it.
Describe alternatives you've considered
Measure the parent first (
info.get(parent).bounds), compute the cell count, and feed an absolute number back on a second pass—but that needs two renders, goes stale on resize, and can't express floor-plus-content-grows in one pass. The whole-axispercent()is the existing primitive that falls short here.Additional context
Failing test case on
nm/repro/percent-sizing(test · diff). Whichever shape wins, the clamp lives deepest in the layout engine:Clay_SizingMinMax(clay/clay.h:316-328) is absolutefloatcells with no parent-relative bound, so it needs a percent-aware min/max on the wire and in the engine.