Indicator
Summary
The Indicator concept is for objects that compute running statistics over a stream of incoming and outgoing samples. Incoming and outgoing samples can be presented individually, or collectively in the form of an object of the same Indicator type.
Refinement of
CopyAssignable, Destructible
Associated types
The sample_type of an Indicator is the type of the samples over which the statistic is calculated.
The weight_type of an Indicator is the type by which samples are weighted.
The result_type is the type of the calculated statistic over the samples.
Notation
indicator_type is a class that models the Indicator concept.
indicator and other are both objects of the type indicator_type.
v is an object of value_type,
w is an object of weight_type,
r is an object of result_type.
Definitions
n.a.
Valid expressions
indicator.add(v, w);
indicator.add(v);
indicator.subtract(v, w);
indicator.subtract(v);
indicator.add(other);
indicator.add(other, w);
indicator.subtract(other);
indicator.subtract(other, w);
r = indicator.extract();
using value_type = typename indicator_traits<indicator_type>::value_type;
using weight_type = typename indicator_traits<indicator_type>::weight_type;
using result_type = typename indicator_traits<indicator_type>::result_type;
Expression Semantics
|Expression|Semantic|
|———-|——–|
|indicator.add(v,w)|update the indicator with an additional sample of value v and weight w|
|indicator.add(v)|update the indicator with an additional sample of value v and unity weight|
|indicator.subtract(v,w)|update the indicator by undoing the effect of indicator.add(v,w)|
|indicator.subtract(v)|update the indicator by undoing the effect of indicator.add(v)|
|indicator.add(other,w)|update the indicator to include the effect of all samples reflected in other, applying a further weight w|
|indicator.add(other)|update the indicator to include the effect of all samples reflected in other|
|indicator.subtract(other,w)|update the indicator by undoing the effect of indicator.add(other,w)|
|indicator.subtract(other)|update the indicator by undoing the effect of indicator.add(other)|
|r = indicator.extract()|get the value of the statistic for the samples currently in the indicator|
Complexity guarantees
n.a.
Invariants
Conceptually, operations must be retractable (subtracting an earlier added sample is equivalent to not having added the observation at all) and communicative (the order of subtracting and adding does not affect results). However, differences may occur due to rounding when floating point numbers are involved. Samples should not be retracted that have not been added before.
Models
The Pronto Raster library includes a number of classes that model Indicator in the include/pronto/raster/indicator/ directory. However, these have not been documented yet.
Notes
In many cases the result_type will be some optional<T> to allow for the circumstance that there are insufficient samples to extract the indicator.