Commit 2026-11-13 15:19 28fa0f19

View on Github →

feat(MeasureTheory): one-sided continuous functions are measurable (#42164) The following lemmas are proved in this PR:

  • IsRightContinuousOn.separableSpace_image, IsRightContinuous.separableSpace_range: the image of a set of a second-countable linear order under a right-continuous function is a separable space.
  • IsRightContinuous.measurable: a right continuous function is measurable.
  • IsRightContinuous.stronglyMeasurable: a right continuous function is strongly measurable. The corresponding statements for left continuous functions are also included, via to_dual. Note: the statement that a right continuous function is strongly measurable is actually true even in the case that the codomain is not assumed to be pseudometrizable, and this can be proved through an explicit construction of an approximating sequence of simple functions. However, this proof is definitely more lengthy, and I believe in practice we only care about strongly measurable functions in the case of integrating banach space valued functions. This is why I decide to assume pseudometrizability so that a simpler proof is available through the use of stronglyMeasurable_iff_measurable_separable. Importing Topology/Order/Cadlag.lean into BorelSpace/Order.lean pulled Analysis.Normed.Group.Continuity into the import closure of Lebesgue/Basic.lean, breaking its assert_not_exists Norm. Cadlag.lean uses no norms, so its import is replaced by Topology.Algebra.GroupWithZero and Topology.MetricSpace.Bounded. That import had also been transitively supplying the push ← attribute on Function.comp_def that @[to_fun] needs to eta-expand g ∘ f, so the file now declares it itself, as Order/ScottContinuity.lean does (see #40183).

Estimated changes