Category: Uncategorized

What will we do in this tutorial?

Create a mosaic in Excel using Python.

See the tutorial on YouTube and hear how you can use it as prank!

Step 1: How to create a mosaic from a photo

If you want a deeper description of how to create a mosaic you should read the following tutorial, which shows the code on how to do it.

Step 2: Install the necessary libraries

You need to install OpenCV. If you use PyCharm you can follow this tutorial.

Otherwise you can install the libraries as follows.

pip install opencv-python
pip install numpy
pip install xlsxwriter

Please ask if you have troubles with it.

Step 3: The code used to create the Mosaic in Excel

This will just provide the code for you to enjoy. The process code is used from the tutorial linked above, where it is described.

import cv2
import numpy as np
import xlsxwriter

def create_mosaic_in_excel(photo, box_height, box_width, col_width=2, row_height=15):
    # Get the height and width of the photo
    height, width, _ = photo.shape
    # Create Excel workbook and worksheet
    workbook = xlsxwriter.Workbook('mosaic.xlsx')
    worksheet = workbook.add_worksheet("Urgent")
    # Resize columns and rows
    worksheet.set_column(0, width//box_width - 1, col_width)
    for i in range(height//box_height):
        worksheet.set_row(i, row_height)
    # Create mosaic
    for i in range(0, height, box_height):
        for j in range(0, width, box_width):
            # Create region of interest (ROI)
            roi = photo[i:i + box_height, j:j + box_width]
            # Use numpy to calculate mean in ROI of color channels
            b_mean = np.mean(roi[:, :, 0])
            g_mean = np.mean(roi[:, :, 1])
            r_mean = np.mean(roi[:, :, 2])
            # Convert mean to int
            b_mean_int = b_mean.astype(int).item()
            g_mean_int = g_mean.astype(int).item()
            r_mean_int = r_mean.astype(int).item()
            # Create color code
            color = '#{:02x}{:02x}{:02x}'.format(r_mean_int, g_mean_int, b_mean_int)
            # Add color code to cell
            cell_format = workbook.add_format()
            cell_format.set_bg_color(color)
            worksheet.write(i//box_height, j//box_width, "", cell_format)
    # Close and write the Excel sheet
    workbook.close()

def main():
    photo = cv2.imread("rune.png")
    number_cols = 50
    number_rows = 45
    # Get height and width of photo
    height, width, _ = photo.shape
    box_width = width // number_cols
    box_height = height // number_rows
    # To make sure that it we can slice the photo in box-sizes
    width = (width // box_width) * box_width
    height = (height // box_height) * box_height
    photo = cv2.resize(photo, (width, height))
    # Create the Excel mosaic
    create_mosaic_in_excel(photo.copy(), box_height, box_width, col_width=2, row_height=15)

main()

Step 4: What to modify

The above tutorial assumes a photo of me in rune.png. I used the one taken from this page. You should obviously change it to something else.

You can change how many columns and rows in the Excel sheet the mosaic should be. This is done by changing the values of number_cols and number_rows.

Then you can change the values of col_width=2 and row_height=15.

In the YouTube video I use this free picture from Pexels (download) and modify number_cols = 100 and number_rows = 90, and col_width=1 and row_height=6.

What will we cover in this tutorial?

The key to use exceptions correct is to understand why we use them.

The best way to understand why to use exceptions is to see what happens when we do not use exceptions in our code.

After that exploration we will show how it can solve the examples with exceptions.

Step 1: In the perfect world!

Remember those times before the exception was invented?

No? Well, of course not. It is long time ago, in a distant past in a programming language you probably have not heard about.

As an aspiring programmer, you have probably seen and heard about exceptions, but never put them to use yourself.

The best way is to create examples that can be greatly simplified with exceptions.

Let’s keep the world simple.

def add(a, b):
    return a + b

print(add(2, 3))

This would print out the result 5.

As long as the world is clean and everybody uses the function correct, there is nothing to worry about.

But what if someone calls the function with wrong types of arguments. Then the function does not do as expected.

def add(a, b):
    return a + b

print(add("2", "3"))

This will print 23, which might be a bit surprising if you are not familiar with the function, and possibly how Python uses addition on strings.

So how to handle it.

Step 2: In the real world where people do not use things as intended

As you will realize in your careers of programming, it happens that other programmers do not use your function correctly.

Why would they do that?

Maybe they don’t take the time to read the good documentation you wrote. Or maybe they are just careless and working under hard time pressure.

Even more funny, it could be you. It could be code you wrote a year ago (or less), and the documentation was not that good as you thought, hence, you use your own function incorrectly.

To continue the simple example. One way to handle wrong input without exceptions could be as follows.

def add(a, b):
    if type(a) is not int or type(b) is not int:
        return None
    return a + b

result = add("2", "3")
if result is not None:
    print(result)
else:
    print("Invalid input format to function add!")

Oh, no! What happened. Actually, your function is not that difficult to understand. It just starts by checking if the input format is as expected. If not, return an error code. As this is Python, an error code can simply be the value None.

That might seem fine. But what about the user of your function? Now the user needs to validate that the correct format of the result was returned.

This makes the code more complex. Also, the user of your function needs to know more details of how your function handles invalid input, or whatever can go wrong in your function.

The problem is, that you pass on a problem to someone else, which needs to know about the details of how your function works.

This is the opposite of decoupling. And decoupling is considered good. Easier to program, easier to change, easier to maintain, to mention a few benefits.

Step 3: Solve it with exceptions

The core idea of exceptions is to handle the when something out of the ordinary happens from the main logic of a program.

What? Yes, the expected case in our simple function, is that the user provides integers as input. But it might happen they do not call it with integers. That is unexpected, that is something out of the ordinary, and this breaks the logic in how you would build the program.

So how should we do in the above example?

def add(a, b):
    if type(a) is not int or type(b) is not int:
        raise Exception("Input format incorrect")
    return a + b

Actually, your part of the obligations becomes easier.

How?

Because now you do not need to write complex documentation of the error return codes in your program. The documentation is kept automatically in the exception. It will even guide the user to where in the code the exception comes from.

def add(a, b):
    if type(a) is not int or type(b) is not int:
        raise Exception("Input format incorrect")
    return a + b

result = add("2", "3")

This would give the following result.

Traceback (most recent call last):
  File "/Users/admin/PycharmProjects/LearningSpace/test2.py", line 7, in <module>
    result = add("2", "3")
  File "/Users/admin/PycharmProjects/LearningSpace/test2.py", line 3, in add
    raise Exception("Input format incorrect")
Exception: Input format incorrect

Which leads the user of the function to where in the code it went wrong. Then it will be easier for them to figure out what is wrong and how to fix it.

Step 4: Try-catch from the user

It happens that the user of your function would like to handle the special case.

They now master what is going wrong, and want to inform their users of the problem.

def add(a, b):
    if type(a) is not int or type(b) is not int:
        raise Exception("Input format incorrect")
    return a + b

try:
    result = add("2", "3")
except:
    print("The input format is incorrect")

This leads to something you might not notice at first. It leads to a good easy flow in your program. It is easy to read and understand.

It is clear that this is not intended flow in the except-part of the program. Also, the normal flow in the program would be in the try-part.

It makes it easier to understand, which is the ultimate goal of a good programmer. Make your code easy to read for others. That is, including you in less than 6 months (we do forget our own code and the logic in the programs we write faster than we expect).

What will we cover in this tutorial?

Wonder why some of these trivial questions are popular in coding interviews?

The truths is, they have a depth of hidden aspects which uncovers whether you understand the the true nature of the problem.

Often, there is no right and wrong solution. It depends. Can you answer what it depends on?

In this tutorial we will investigate the Anagram Problem.

What they are really looking for is your underlying understanding of algorithmic time complexity (big-O analysis) and your ability to improve on solutions.

This tutorial will show one example of this.

Step 1: What is the Anagram Problem?

The first part of a coding challenge problem is to show that you understand it. Many underestimate this part and do not realize how important that is.

Often the problem statements are a bit vague, but given with some examples. It could be directly taken from wikipedia and ask you to implement it.

An anagram is a word or phrase formed by rearranging the letters of a different word or phrase.

https://en.wikipedia.org/wiki/Anagram

Or as it could be stated in you interview question: Given two strings, str_a and str_b, return whether they are an anagram of each other.

First, you need to understand the problem. Often there will be examples to show you what an anagram is.

An examples could include.

• “Listen” and “Silent”

Where you notice that the letters in LISTEN are the exact same as in SILENT, just in a different order.

What you also notice is that the lowercase and uppercase of the letters do not matter.

The definition says also phrases, hence, examples of anagrams include.

• “a gentleman” and “elegant man”
• “eleven plus two” and “twelve plus one”
• “William Shakespeare” = “I am a weakish speller”

The last example is quite interesting, as it shows that the number of spaces do not seem to matter. The phrase “William Shakespeare” has 1 space, while the phrase “I am a weakish speller” has 4 spaces.

Now your part is to show that you understand it. In this case it would be good to state an input that will return True and one that will return False.

Say, “evil” and “vile” as input returns True, while “evil” and “test” will return False.

If you do not understand it, don’t worry. Imagine that you do not admit it. Then what is the best that can happen? Are you by any chance going to guess it out of the blue? Most likely not.

The best you can do is to admit it and explain as good as you can as much of what you think you understand.

Step 2: Find edge cases and set restrictions on input

A natural next step is to consider edge cases and set restrictions. Most would jump into writing some pseudo code and try to solve it. Don’t be that type of person. Take a deep breath and remember what we said in Step 1.

What did we observe in Step 1?

1. That uppercase and lowercase is ignored when considering anagrams.
2. That the number of spaces do not matter when considering anagrams.

Now this is important to notice.

An advise is to simplify as much as possible in a job interview. But you need to tell them that you do simplify the problem.

Hence, you can say.

1. You assume that input will be all in lowercase.
2. That input will not have spaces.

Is this reasonable? Well, it simplifies your code a lot. If they want you to consider these cases, they will let you know.

The importance of this step cannot be stressed enough. Because, if you just assumes that you can restrict input and write your code, they might now you have considered theses cases.

On the other hand, if you do set it as restrictions before hand they know you have thought of it.

Step 3: The naive solution

The next step is to make a solution. As you might not come up with the best solution first, a good way is to start with whatever comes to your mind and call it the naive solution.

It is always a good idea to write it down. This will help you get a better idea to solve it more efficient afterwards. Notice, that the restrictions you made in Step 2 comes in handy now, as you can construct an easy solution fast now.

How could a naive solution look like?

def is_anagram_naive(str_a, str_b):
    if len(str_a) != len(str_b):
        return False
    for c in str_a:
        if c not in str_b:
            return False
        else:
            str_b = str_b.replace(c, '', 1)
    return True

print(is_anagram_naive("silent", "listen"))
print(is_anagram_naive("elevenplustwo", "twelveplusone"))
print(is_anagram_naive("anna", "anaa"))

The question is how does it perform?

Well, first the check of the length of the string should be in constant time, O(1). If the underlying interpreter does not have an efficient implementation, then it could be O(n) time, where n is the length of str_a and str_b.

The loop over str_a is O(n). In each loop it will check if the character c is in str_b. This check is implemented to be faster than brute force. In an job interview you are not expected to remember that. It is good enough to say O(n) in first iteration.

Then we notice that it will use n, n-1, n-2, …, 3, 2, 1 operations in total to do this test. This results to n(n-1)/2 operations, which is O(n^2) in total.

On top of that, it will use the same to replace the character c in str_b.

That means the algorithm uses O(n^2) time complexity in worst case.

Step 3: Improving the performance

This is where it gets interesting. Now you showed you can solve the problem. The next step is to show that you can improve it.

How to get an idea?

Good you asked.

It lies in the previous step. What is the time complexity. You need to improve on the performance. What is often better than O(n^2)?

Yes, a natural next improvement would be O(n log(n)). What algorithms do you know? Well, sorting, can that help?

What if we sort the characters and compare them in order, then they must be identical if it is an anagram. Try to sort the characters of any anagram. Example, “silent” becomes [‘e’, ‘i’, ‘l’, ‘n’, ‘s’, ‘t’], and “listen” becomes [‘e’, ‘i’, ‘l’, ‘n’, ‘s’, ‘t’], identical.

This makes sense. As they have the same characters, they must come in the same order if we have them sorted.

Let’s try.

def is_anagram_improved(str_a, str_b):
    if len(str_a) != len(str_b):
        return False
    list_a = list(str_a)
    list_b = list(str_b)
    list_a.sort()
    list_b.sort()
    for l1, l2 in zip(list_a, list_b):
        if l1 != l2:
            return False
    return True

We start by checking if the length of the strings are the same, if not, return False, as it cannot be anagrams.

Then convert the strings to lists. It should take O(n) time. Now for the expensive step. Sorting it. This takes O(n log(n)) time.

Finally we check the characters one by one, in sorted order, if they are equal. If not, we do not have an anagram. This takes O(n) time.

That is, it uses O(n) + O(n log(n)) + O(n) = O(n log(n)) time.

You did the first improvement.

Can you do better? I know, this is annoying. But maybe we can. Let’s try.

Step 4: Improving the code – a common pitfall

Well, you are a shark at Python. Hence you know you that the above can be made into a more Pythonic code.

You feel on fire.

Yes, this can be done smarter, you say.

def is_anagram_pythonic_way(str_a, str_b):
    return sorted(str_a) == sorted(str_b)

It does actually do the same and in only one line of code. Isn’t that what coding is all about. You showcase how good you are and how much you know?

Well, both yes an no.

Yes, because it does show you are familiar with coding in Python.

No, because it did not improve the algorithmic time complexity. It still runs in O(n log(n)).

So what is gained? What did you tell the interviewer?

They are looking for improvement of the algorithmic time complexity or space complexity. Hence, showing them a smarter way to write the code does not add many points to the book.

I know it feels awesome to write a one-liner that does the same, but that is not what they are looking for.

The advice is, limit your time on improving lines of code, if it does not improve the algorithmic time complexity or space complexity.

Step 5: Improving the performance again

There is often a even better way to solve this things.

How do you get an idea to do that? Well, look at what takes time in the algorithm from Step 3. It is the sorting. Without the sorting step, you could have an algorithm in O(n) time.

Can we do something different than sorting?

What does the sorting do for us? It arranges the characters in order, repeating each character with the number of occurrences.

What if we had a table and just updated the number of occurrences of each character?

Let’s try.

def is_anagram_further_improvements(str_a, str_b):
    occurrences_a = [0]*256
    occurrences_b = [0]*256
    for c in str_a:
        occurrences_a[ord(c)] += 1
    for c in str_b:
        occurrences_b[ord(c)] += 1
    for i1, i2 in zip(occurrences_a, occurrences_b):
        if i1 != i2:
            return False
    return True

First notice we just assume 256 characters, which is not needed for the assumption. It just simplifies some calculations of the characters. In a job interview they are not looking for the optimal final solution, just if you understand the concepts of big-O time complexity.

The initialization of the size 256 occurrences counters is independent of the size, hence O(1) time.

Then the following two for-loops take O(n) time. Finally, the last for-loop takes O(256) = O(1) time.

This is a total of O(n) time complexity of the algorithm.

See, we could do some awesome stuff together.

Conclusion

Is this the perfect written solution? No, this is a job interview. Time is limited. It is important to show what they are looking for.

Do you understand big-O time complexity. Do you see how a simple solution can be improved?

The anagram is one of the favorite coding interview problems, because it has this natural flow of improving it. They do not expect you to remember the optimal solution. But they expect you to reason your way forward as we did in this tutorial.