firanka is a Python library to perform calculations on particular kinds of functions. These functions have a domain, which is a single continuous subset of the real number line. These functions can have any values.
firanka allows you do define two classes of such functions or series.
First are the DiscreteSeries. DiscreteSeries further divide the function domain into slices (left-closed, right-open) that have constant values. Manipulating DiscreteSeries and performing calculations on them is cheap.
Then you have FunctionSeries. These are simply defined by user-supplied Python callable.
Best part is, you can join series together (given a joining operator), slice them and so on.
Can be imported from sai.series. A generic abstract superclass for series -
Series can be imported for checking if given object is a series.
Series are immutable, but non-hashable.
Read the source code of the base class to get to know more about series operations.
Applying and joining
Applying requires a callable(index: float, value: current value) -> value. Joining requires a callable(index: float, valueSelf, valueOther: values from self and other table) -> value.
To use a DiscreteSeries you must give it a set of data to work with. These will define intervals with given values, left-closed, right-open. as in:
fs = DiscreteSeries([(0,1), (3, 4), (5, 6)]) fs[0.5] == 1 fs == 4 fs == 6 fs.domain == '<0;5>' # fs - NotInDomainError's
Datapoints given must be already sorted!. By default, the domain will be both sides closed, from minimum to maximum given in data, but you can specify a custom one:
fs = DiscreteSeries([(0,1), (3, 4), (5, 6)], '(0; 8>') # fs - NotInDomainError's ! fs == 6
Although you can't specify a domain where it would be impossible to compute the value. (ie. starting at smaller than zero). Doing so will throw a ValueError.
Note that when using
join_discrete() sometimes other series might get calls
from beyond their domain. This can be seen for example here:
logs = FunctionSeries(math.log, '(0;5>') dirs = DiscreteSeries([(0,1)], '<0;5>') # Raises ValueError due to math.log being called with 0 dirs.join_discrete(logs, lambda x, y: x+y)
Using FunctionSeries is straightforward. Just give them a callable and a domain:
fs = FunctionSeries(lambda x: x**2, '<-2;2>')
ModuloSeries allow you to wrap a finite series in repetition.
fs = ModuloSeries(someOtherSeries)
By definition, ModuloSeries has the domain of all real numbers.
Note that someOtherSeries's domain length must be non-zero and finite. Otherwise ValueError will be thrown.
These are discretes, but allow you to define an operator that will take its neighbours into account and let you return a custom value.
By default, it will assumes that values can be added, subbed, multed and dived, and will do classical linear interpolation.
They can either utilize an existing discrete series, or be created just as any other discrete series would be.
Sometimes you just need to update a DiscreteSeries, or to blang a brand new one. This little fella will help you out.
You can pass a DiscreteSeries to build on or start from stratch:
kb = DiscreteSeriesBuilder(series) kb = DiscreteSeriesBuilder() kb.put(1,2) series = kb.as_series() isinstance(series, DiscreteSeries)
as_series() you get a new DiscreteSeries instance returned.
Can be imported from sai.intervals.
Interval is a continuous subset of the real number line.
You can create Intervals as follows:
Interval(-5, 5, True, False) == Interval('<-5;5)')
For more information use the source Interval's are immutable and hashable. They can be sliced:
Interval('<-5;5>')[0:] == Interval('<0;5>')
Slices work as a both-sides-closed range if both sides are shown!
You can check whether an interval contains a point
5 not in Interval('<-1;5)')
Or you can check for strict inclusion
Interval('<-1;1>') in Interval('<-2;2>')