mirror of
https://github.com/TheAlgorithms/Python.git
synced 2024-12-05 02:40:16 +00:00
0d324de7ab
* Python mirror_formulae.py is added to the repository * Changes done after reading readme.md * Changes for running doctest on all platforms * [pre-commit.ci] auto fixes from pre-commit.com hooks for more information, see https://pre-commit.ci * Change 2 for Doctests * Changes for doctest 2 * updating DIRECTORY.md * Doctest whitespace error rectification to mirror_formulae.py * updating DIRECTORY.md --------- Co-authored-by: pre-commit-ci[bot] <66853113+pre-commit-ci[bot]@users.noreply.github.com> Co-authored-by: github-actions <${GITHUB_ACTOR}@users.noreply.github.com> Co-authored-by: Christian Clauss <cclauss@me.com>
128 lines
4.9 KiB
Python
128 lines
4.9 KiB
Python
"""
|
|
This module contains the functions to calculate the focal length, object distance
|
|
and image distance of a mirror.
|
|
|
|
The mirror formula is an equation that relates the object distance (u),
|
|
image distance (v), and focal length (f) of a spherical mirror.
|
|
It is commonly used in optics to determine the position and characteristics
|
|
of an image formed by a mirror. It is expressed using the formulae :
|
|
|
|
-------------------
|
|
| 1/f = 1/v + 1/u |
|
|
-------------------
|
|
|
|
Where,
|
|
f = Focal length of the spherical mirror (metre)
|
|
v = Image distance from the mirror (metre)
|
|
u = Object distance from the mirror (metre)
|
|
|
|
|
|
The signs of the distances are taken with respect to the sign convention.
|
|
The sign convention is as follows:
|
|
1) Object is always placed to the left of mirror
|
|
2) Distances measured in the direction of the incident ray are positive
|
|
and the distances measured in the direction opposite to that of the incident
|
|
rays are negative.
|
|
3) All distances are measured from the pole of the mirror.
|
|
|
|
|
|
There are a few assumptions that are made while using the mirror formulae.
|
|
They are as follows:
|
|
1) Thin Mirror: The mirror is assumed to be thin, meaning its thickness is
|
|
negligible compared to its radius of curvature. This assumption allows
|
|
us to treat the mirror as a two-dimensional surface.
|
|
2) Spherical Mirror: The mirror is assumed to have a spherical shape. While this
|
|
assumption may not hold exactly for all mirrors, it is a reasonable approximation
|
|
for most practical purposes.
|
|
3) Small Angles: The angles involved in the derivation are assumed to be small.
|
|
This assumption allows us to use the small-angle approximation, where the tangent
|
|
of a small angle is approximately equal to the angle itself. It simplifies the
|
|
calculations and makes the derivation more manageable.
|
|
4) Paraxial Rays: The mirror formula is derived using paraxial rays, which are
|
|
rays that are close to the principal axis and make small angles with it. This
|
|
assumption ensures that the rays are close enough to the principal axis, making the
|
|
calculations more accurate.
|
|
5) Reflection and Refraction Laws: The derivation assumes that the laws of
|
|
reflection and refraction hold.
|
|
These laws state that the angle of incidence is equal to the angle of reflection
|
|
for reflection, and the incident and refracted rays lie in the same plane and
|
|
obey Snell's law for refraction.
|
|
|
|
(Description and Assumptions adapted from
|
|
https://www.collegesearch.in/articles/mirror-formula-derivation)
|
|
|
|
(Sign Convention adapted from
|
|
https://www.toppr.com/ask/content/concept/sign-convention-for-mirrors-210189/)
|
|
|
|
|
|
"""
|
|
|
|
|
|
def focal_length(distance_of_object: float, distance_of_image: float) -> float:
|
|
"""
|
|
>>> from math import isclose
|
|
>>> isclose(focal_length(10, 20), 6.66666666666666)
|
|
True
|
|
>>> from math import isclose
|
|
>>> isclose(focal_length(9.5, 6.7), 3.929012346)
|
|
True
|
|
>>> focal_length(0, 20) # doctest: +NORMALIZE_WHITESPACE
|
|
Traceback (most recent call last):
|
|
...
|
|
ValueError: Invalid inputs. Enter non zero values with respect
|
|
to the sign convention.
|
|
"""
|
|
|
|
if distance_of_object == 0 or distance_of_image == 0:
|
|
raise ValueError(
|
|
"Invalid inputs. Enter non zero values with respect to the sign convention."
|
|
)
|
|
focal_length = 1 / ((1 / distance_of_object) + (1 / distance_of_image))
|
|
return focal_length
|
|
|
|
|
|
def object_distance(focal_length: float, distance_of_image: float) -> float:
|
|
"""
|
|
>>> from math import isclose
|
|
>>> isclose(object_distance(30, 20), -60.0)
|
|
True
|
|
>>> from math import isclose
|
|
>>> isclose(object_distance(10.5, 11.7), 102.375)
|
|
True
|
|
>>> object_distance(90, 0) # doctest: +NORMALIZE_WHITESPACE
|
|
Traceback (most recent call last):
|
|
...
|
|
ValueError: Invalid inputs. Enter non zero values with respect
|
|
to the sign convention.
|
|
"""
|
|
|
|
if distance_of_image == 0 or focal_length == 0:
|
|
raise ValueError(
|
|
"Invalid inputs. Enter non zero values with respect to the sign convention."
|
|
)
|
|
object_distance = 1 / ((1 / focal_length) - (1 / distance_of_image))
|
|
return object_distance
|
|
|
|
|
|
def image_distance(focal_length: float, distance_of_object: float) -> float:
|
|
"""
|
|
>>> from math import isclose
|
|
>>> isclose(image_distance(10, 40), 13.33333333)
|
|
True
|
|
>>> from math import isclose
|
|
>>> isclose(image_distance(1.5, 6.7), 1.932692308)
|
|
True
|
|
>>> image_distance(0, 0) # doctest: +NORMALIZE_WHITESPACE
|
|
Traceback (most recent call last):
|
|
...
|
|
ValueError: Invalid inputs. Enter non zero values with respect
|
|
to the sign convention.
|
|
"""
|
|
|
|
if distance_of_object == 0 or focal_length == 0:
|
|
raise ValueError(
|
|
"Invalid inputs. Enter non zero values with respect to the sign convention."
|
|
)
|
|
image_distance = 1 / ((1 / focal_length) - (1 / distance_of_object))
|
|
return image_distance
|