Skip to content

Constrain essentiallyEqual and essentiallyZero

Gerhard Raven requested to merge constrain-essentiallyEqual-essentiallyZero into master

LHCb::essentiallyEqual and LHCb::essentiallyZero are intended to avoid imprecise floating point comparisons -- but the current version does not actually check whether it is invoked with a floating point (like) type. This MR requires the types of the argument to have std::numeric_limits to be specialized for them (otherwise the use of std::numeric_limits<T>::epsilon() and std::numeric_limits<T>::min() makes no sense), and it distinguishes between 'exact' types (such as integer and bool) and 'non-exact' (such as float and double). Only for the 'inexact' types are the comparisons now done in a way which takes into account that these types are not 'exact'.

(note: yes, essentiallyZero does currently get invoked with eg. a bool argument, and essentiallyEqual is currently invoked with plain int arguments )

Validated by

  • Core Software
  • RTA
  • Simulation
Edited by Sebastien Ponce

Merge request reports