lazy_poly Module

Polynomial model and Waring-Lagrange polynomial interpolator

Summary of module contents:

Name	Description
PolyMeta	Poly metaclass. This class overloads few operators to the Poly class. All binary dunders (non reverse) should be implemented on the Poly class
Poly	Model for a polynomial, a Laurent polynomial or a sum of arbitrary powers.
x	Model for a polynomial, a Laurent polynomial or a sum of arbitrary powers.
lagrange	This is a StrategyDict instance object called lagrange. Strategies stored: 2.
resample	Generic resampler based on Waring-Lagrange interpolators.

class PolyMeta

Bases: audiolazy.lazy_core.AbstractOperatorOverloaderMeta

Poly metaclass. This class overloads few operators to the Poly class. All binary dunders (non reverse) should be implemented on the Poly class

__operators__ = '+ - * pow truediv eq ne '

__rbinary__(op)

__unary__(op)

class Poly(data=None, zero=None)

Bases: object

Model for a polynomial, a Laurent polynomial or a sum of arbitrary powers.

The “values” method allows casting simple polynomials to a list with list(Poly.values()) where the index values are the powers. The “terms” method allows casting any sum of powers to dict with dict(Poly.terms()) where the keys are the powers and the values are the coefficients, and this method can also give the terms sorted by their power value if needed. Both methods return a generator.

Usually the instances of this class should be seen as immutable (this is a hashable instance), although there’s no enforcement for that (and item set is allowed) until the hash is required.

You can use the x object and operators to create your own instances.

Examples:

>>> x ** 5 - x + 7
7 - x + x^5
>>> type(x + 1)
<class 'audiolazy.lazy_poly.Poly'>
>>> (x + 2)(17)
19
>>> (x ** 2 + 2 * x + 1)(2)
9
>>> (x ** 2 + 2 * x + 1)(.5)
2.25
>>> (x ** -2 + x)(10)
10.01
>>> spow = x ** -2.1 + x ** .3 + x ** -1 + x - 6
>>> value = spow(5)
>>> "{:.6f}".format(value) # Just to see the first few digits
'0.854710'

__add__(other)

__call__(value, horner='auto')

Apply the given value to the Poly.

Parameters:

• value – The value to be applied. This can possibly be another Poly, or perhaps a Stream instance.
• horner – A value in [True, False, “auto”] which chooses whether a Horner-like scheme should be used for the evaluation. Defaults to "auto", which means the scheme is done only for simple polynomials. This scheme can be forced on Laurent polynomials and any other sum of powers that have sortable powers. This input is neglect when value is a Poly instance.

Note

For a polynomial with all values, the Horner scheme is done like expected, e.g. replacing x ** 2 + 2 * x + 1 by x * (x + 2) + 1 in the evaluation. However, for zeroed coeffs, it will merge steps, using powers instead of several multiplications, e.g. x ** 7 + x ** 6 + 4 would evaluate as x ** 6 * (x + 1) + 4, taking the x ** 6 instead of multiplying (x + 1) by x six times (i.e., the evaluation order is changed). This allows a faster approach for sparse polynomials and extends the scheme for any sum of powers with sortable powers.

__eq__(other)

__getitem__(item)

__hash__()

__init__(data=None, zero=None)

Inits a polynomial from given data, which can be a list or a dict.

A list \([a_0, a_1, a_2, a_3, ...]\) inits a polynomial like

\[a_0 + a_1 . x + a_2 . x^2 + a_3 . x^3 + ...\]

If data is a dictionary, powers are the keys and the \(a_i\) factors are the values, so negative powers are allowed and you can neglect the zeros in between, i.e., a dict with terms like {power: value} can also be used.

__len__()

Number of terms, not values (be careful).

__mul__(other)

__ne__(other)

__neg__()

__pos__()

__pow__(other)

Power operator. The “other” parameter should be an int (or anything like), but it works with float when the Poly has only one term.

__radd__(other)

__repr__()

__rmul__(other)

__rsub__(other)

__setitem__(power, coeff)

__str__()

__sub__(other)

__truediv__(other)

copy(zero=None)

Returns a Poly instance with the same terms, but as a “T” (tee) copy when they’re Stream instances, allowing maths using a polynomial more than once.

diff(n=1)

Differentiate (n-th derivative, where the default n is 1).

integrate()

Integrate without adding an integration constant.

is_laurent()

Boolean that indicates whether is a Laurent polynomial or not.

A Laurent polynomial is any sum of integer powers of x.

Examples:

>>> (x + 4).is_laurent()
True
>>> (x ** -3 + 4).is_laurent()
True
>>> (x ** -3 + 4).is_polynomial()
False
>>> (x ** 1.1 + 4).is_laurent()
False

is_polynomial()

Tells whether it is a linear combination of natural powers of x.

order

Finds the polynomial order.

Examples:

>>> (x + 4).order
1
>>> (x + 4 - x ** 18).order
18
>>> (x - x).order
0
>>> (x ** -3 + 4).order
Traceback (most recent call last):
  ...
AttributeError: Power needs to be positive integers

roots

Returns a list with all roots. Needs Numpy.

terms(sort='auto', reverse=False)

Pairs (2-tuple) generator where each tuple has a (power, value) term, perhaps sorted by power. Useful for casting as dict.

Parameters:

• sort – A boolean value or "auto" (default) which chooses whether the terms should be sorted. The "auto" value means True for Laurent polynomials (i.e., integer powers), False otherwise. If sorting is disabled, this method will yield the terms in the creation order.
• reverse – Boolean to chooses whether the [sorted or creation] order should be reversed when yielding the pairs. If False (default), yields in ascending or creation order (not counting possible updates in the coefficients).

values()

Array values generator for powers from zero to upper power. Useful to cast as list/tuple and for numpy/scipy integration (be careful: numpy use the reversed from the output of this function used as input to a list or a tuple constructor).

zero

resample(sig, old=1, new=1, order=3, zero=0.0)

Generic resampler based on Waring-Lagrange interpolators.

Parameters:

• sig – Input signal (any iterable).
• old – Time duration reference (defaults to 1, allowing percentages to the new keyword argument). This can be float number, or perhaps a Stream instance.
• new – Time duration that the reference will have after resampling. For example, if old = 1, new = 2, then there will be 2 samples yielded for each sample from input. This can be a float number, or perhaps a Stream instance.
• order – Lagrange interpolator order. The amount of neighboring samples to be used by the interpolator is order + 1.
• zero – The input should be thought as zero-padded from the left with this value.

Returns:

The first value will be the first sample from sig, and then the interpolator will find the next samples towards the end of the sig. The actual sampling interval (or time step) for this interpolator obeys to the old / new relationship.

Hint

The time step can also be time-varying, although that’s certainly difficult to synchonize (one sample is needed for each output sample). Perhaps the best approach for this case would be a ControlStream keeping the desired value at any time.

Note

The input isn’t zero-padded at right. It means that the last output will be one with interpolated with known data. For endless inputs that’s ok, this makes no difference, but for finite inputs that may be undesirable.

lazy_poly.lagrange StrategyDict

This is a StrategyDict instance object called lagrange. Strategies stored: 2.

Strategy lagrange.func (Default). Docstring starts with:

Waring-Lagrange interpolator function.

Strategy lagrange.poly. Docstring starts with:

Waring-Lagrange interpolator polynomial.

Note

This docstring is self-generated, see the StrategyDict class and the strategies docs for more details.

poly(pairs)

Waring-Lagrange interpolator polynomial.

Parameters:pairs – Iterable with pairs (tuples with two values), corresponding to points (x, y) of the function. Returns:A Poly instance that allows finding the interpolated value for any x.

func(pairs)

Waring-Lagrange interpolator function.

Parameters:pairs – Iterable with pairs (tuples with two values), corresponding to points (x, y) of the function. Returns:A function that returns the interpolator result for a given x.