Longest Repeating Character Replacement - Leetcode 424 - Python

Поделиться
HTML-код
  • Опубликовано: 7 янв 2025

Комментарии • 361

  • @NeetCode
    @NeetCode  2 года назад +71

    🚀 neetcode.io/ - I created a FREE site to make interview prep a lot easier, hope it helps! ❤

    • @codeguru4
      @codeguru4 Год назад

      hi Neetcode In you Website Code Submission isn't auto Updating we need to refresh inoreder to get the latest submission

  • @elheffe2597
    @elheffe2597 3 года назад +273

    I love that you always start with the most obvious brute-force solution first, and then show how to optimize it after. It makes these problems so much easier to digest.

    • @aznguyener
      @aznguyener Год назад +4

      Also for interviewees, its good to talk this through out loud in an interview so they can see your thought process!

  • @jeremyyeo7092
    @jeremyyeo7092 Год назад +281

    For those who are struggling to understand the optimisation with maxf, here is how i understood it:
    For a substring to be valid, we need window_length - maxf

    • @_dion_
      @_dion_ Год назад +15

      probably the best possible breakdown. thank you, jeremy.

    • @JamesBond-mq7pd
      @JamesBond-mq7pd Год назад +6

      he said same thing at 14:00
      "the main idea is this ..."

    • @TaikiTanaka-n6c
      @TaikiTanaka-n6c 10 месяцев назад +18

      @@JamesBond-mq7pd I came to the comments because I was struggling to understand Neatcode's explanation of why we don't need to update the max frequency. The comment above explains it in a much clearer way IMO

    • @praneethkaturi9321
      @praneethkaturi9321 9 месяцев назад +2

      @@TaikiTanaka-n6cI agree. The wording here is better!

    • @ichigo9688
      @ichigo9688 8 месяцев назад

      @neetcode needs to pin this comment.

  • @kartiksoni825
    @kartiksoni825 2 года назад +50

    Please keep doing this for more Leetcode problems - nobody explains them like you do. This channel deserves so many more subscribers!

  • @vishtree
    @vishtree Год назад +7

    Its easier to understand the final optimisation, if you simply replace while condition with if condition. In that case, it simply means that if window_length - maxf > k, its time to shift the window by one position to the right, because you have exceeded the number of replacement you can do in current window.

    • @michaelmarchese5865
      @michaelmarchese5865 10 месяцев назад +1

      Great insight. The while loop was unnecessary all along and just overcomplicated both the overall solution and the maxf optimisation.

  • @votanlean
    @votanlean 9 месяцев назад +4

    Brute Force:
    ans = 0
    for i in range(len(s)):
    for j in range(i, len(s)):
    charCnt = Counter(s[i:j+1])
    for cnt in charCnt.values():
    if j - i + 1 - cnt

  • @jinyang4796
    @jinyang4796 2 года назад +14

    Thank you so much for this video! Loved the part from 12:10 onwards.
    Just adding on, I realised one possible small tweak is that we can use an if condition, instead of a while condition. Saw the following in one of the LC's comments and felt it is really enlightening: "We have a "longest" window already so finding another one of the same size is not helpful. Yeah, we might shrink the window, then extend it, then find another window of the same size as longest but in the end we will see no improvement as long as it it's just as long as longest."

    • @NeetCode
      @NeetCode  2 года назад +2

      Good point

    • @rosenoire9670
      @rosenoire9670 2 года назад +1

      thanks! I was stuck at this little detail as well... although as with other LC problems, it still does not click 100% for me lol.

    • @del6553
      @del6553 10 месяцев назад +1

      If current substring is not valid, then there exist a previous valid substring of currSize-1 (you can prove this by contradiction for example)
      Which implies the final answer >= currSize-1, so now we're only interested in finding res >= currSize
      Therefore we only have to shift left pointer by 1, effectively shifting the current window of currSize by one when we're in the next iteration

    • @manasisingh294
      @manasisingh294 4 месяца назад +1

      Right, as the question doesn't really specify the first or the last longest substring, hence it doesn't really matter as long as the substring is the largest. Our code will always likely give us the last longest substring even if a 'longest' substring of that length was already found at an earlier stage.

  • @ladyking83
    @ladyking83 2 года назад +97

    it's not easy to get one's head around this one, thanks for working it through!💚

    • @lifeofme3172
      @lifeofme3172 Год назад

      True, thanks @Neetcode that slow explanation really helped my brain

    • @Axl124124
      @Axl124124 Год назад

      yeah this one is tough to think about.

  • @kristofferpanahon9913
    @kristofferpanahon9913 3 года назад +110

    You're the best out there - thank you for everything!

  • @joydeeprony89
    @joydeeprony89 2 года назад +2

    there are many channel who has created the content for this problem but you only have explained why do we no need to update maxF in while loop. No fake knowledge , you are pure talent.

  • @jessiz-
    @jessiz- 2 года назад +9

    Having the visualization was really helpful and now the problem seems much simpler. I like how you also explained the max frequency count optimization and the logic behind it. Thanks for your videos!

  • @SandeepKumar16
    @SandeepKumar16 2 года назад +39

    If someone is wondering like me - "Why is while loop and if statement, both giving correct result?"
    The answer is: Suppose we use while loop. It will enter the while loop, only when (strLength - maxFreq) is just greater than k by " 1 ". And as soon as this happens, we decrease our String size by 1, by shifting the left pointer. So, now strLength - maxFreq = k. So, while condition will break.
    In short- While condition will be satisfied only once. So, why not use an "if" instead. Hope this helps someone.
    Thanks @NeetCode. You've done a great job.

    • @АндрейВласов-д3ь
      @АндрейВласов-д3ь Год назад

      thank you, that's exactly what i was looking for in comments, i did wonder why my if statement worked actually

    • @allenlam5210
      @allenlam5210 Год назад

      thank you

    • @jeremyyeo7092
      @jeremyyeo7092 Год назад +11

      In response to Sandeep's comment about the while loop:
      I believe there's a misunderstanding. The condition (window_length - highest_freq) > k can be satisfied more than once for a given window. Consider the example s="ABABACB", k=2:
      We expand the window until it reads "ABABAC".
      Now, let's evaluate the condition:
      "ABABAC"
      window_size=6
      max_freq=3
      window_size−max_freq=3, which is greater than 2.
      So, the condition is satisfied the first time, and we advance the left pointer resulting in "BABAC".
      Re-evaluate:
      "BABAC"
      window_size=5
      max_freq=2
      window_size−max_freq=3, still greater than 2.
      The condition is satisfied for the second time, and we advance the left pointer to get "ABAC".
      Now, the condition is no longer satisfied.
      From this example, we see that the condition can be satisfied multiple times, contrary to the claim that it will only be satisfied once.
      The real reason an "if" statement suffices is that even if the resulting window after the "if" condition is executed might still be invalid, its size will never surpass the maximum window size found up to that point. We are only shrinking the window by 1 by moving the left pointer while keeping the right pointer static. This ensures the correctness of the solution.

    • @fujiantao7680
      @fujiantao7680 Год назад +4

      I think it should make it more clear on why we could move right point or left point, when "window_size−max_freq" is larger than K, moving right point will get no chance to make "window_size−max_freq" smaller, either both window_size + 1 and max_freq + 1, or just window_size + 1 and max_freq keeps the same, while moving left point will reduce window_size by 1, while max_freq could be the same or max_freq-1, that said, moving left will get a chance to make "window_size−max_freq" smaller, that it why when "window_size−max_freq" is larger than K, we just need to move left pointer.
      However, since we are finding the maximum result, so there is no chance to make it better if we shrink the window by pop left and keep right the same, we have to pop left and then move the right and try to see if we can find better solution. However if we just pop left once as the example above, and still not able to get "window_size−max_freq

    • @HoangTran-kf1ij
      @HoangTran-kf1ij Год назад

      thanks you🥰@@fujiantao7680

  • @triscuit5103
    @triscuit5103 3 года назад +6

    fantastic explanation, the first time I get very clearly why we don't need to decrement the maxF var. much love to you, some hugs and kisses as well, but in a very professional and thankful manner.

  • @n1724k
    @n1724k 5 месяцев назад +2

    After listening to your second solution, I coded it before looking at your code, and got completely same one. Kind of proud about it!

  • @whitest_kyle
    @whitest_kyle 2 года назад +54

    This is one of those problems that literally makes no sense to me, like why would we ever need to do this? Tech hiring is so weird...

  • @vudat1710
    @vudat1710 3 года назад +13

    the best of the best. I'm learning a lot from you. Huge credit to you my guy!

  • @dorondavid4698
    @dorondavid4698 3 года назад +22

    You could also stop the algorithm if the R is at the furthest right as when you shift over the Left pointer; if r-l+1

    • @jawakar8266
      @jawakar8266 7 месяцев назад +1

      This doesn't work when s='ABAA' k=0, the result should be 2 but you will get 1

    • @manasisingh294
      @manasisingh294 4 месяца назад +1

      nah, won't work in every case.

  • @Mingfang-k5f
    @Mingfang-k5f Месяц назад +2

    Here is how I understood the optimized solution:
    The problem with not shrinking maxf is that, it is possible to consider an invalid substring which needs k+1 replacements as a valid substring. And if our result comes from the length of an invalid substring we're doomed.
    But the lucky thing is, when we need to shrink the sliding window and shrink maxf, our most recent result must be the substring that barely satisfied the constraint "length

  • @ijavd
    @ijavd 2 года назад +8

    My personal notes on why the O(n) solution works:
    This is our equation: length - maxFreq

    • @qwertyasdf6301
      @qwertyasdf6301 2 года назад +4

      The reason you have given for why the maxfreq need not be updated in case of a decrease is that the result doesn't change. Which is understandable because the length of the substring did not increase.
      But how to answer this:
      Not updating the maxfreq will mess up the valid string window and subsequent manipulations of the window?
      There's no explanation on why the subsequent manipulations of the string window will still be correct given the maxf doesn't hold the accurate value.

    • @publich
      @publich Год назад +1

      @@qwertyasdf6301 I have the same question, did you figure out why not updating maxf doesn't mess up the window manipulations?

    • @h3ckphy246
      @h3ckphy246 6 месяцев назад +1

      @@qwertyasdf6301 "Not updating the maxfreq will mess up the valid string window and subsequent manipulations of the window?"
      That's not true. It won't mess it up. The algorithm works the same. Just take an example string and manually perform all the actions with and without the optimization. You will see that sliding window are the same in each case and on each step.
      Personally I tried this string: AAABBCCDAA.
      Your window will grow to the size of 6 and keep sliding until it reaches the end.
      It's hard to describe everything in the comment sections without illustrations, but drawing with pen and paper helped me to understand it. You (and the others reading this) should do the same.

    • @abdebug
      @abdebug 4 месяца назад

      @@h3ckphy246 "Personally I tried this string: AAABBCCDAA." And what is your value of k for this example ? I will try it out.

  • @madhavoberoi6441
    @madhavoberoi6441 3 года назад +14

    Thanks a lot for this solution, I came up with a bit complex recursive + DP solution with O(n*n) complexity and was proud of myself unless I saw your solution.
    Also keep up the great work you're doing!

  • @anmolpansari9817
    @anmolpansari9817 3 года назад +3

    Loved the second approach.. Optimized Channel - No complexity in finding good videos .. this defines NeetCode

  • @aryanyadav3926
    @aryanyadav3926 2 года назад +5

    Well explained!
    The O(n) solution was really tricky.

  • @AkshayAmbekar-kd8zm
    @AkshayAmbekar-kd8zm Год назад +1

    The optimization with maxf is so freaking genius! Thanks for taking time to make this. Grateful🙏

  • @MaxFung
    @MaxFung 10 месяцев назад

    This explanation is far better than Leetcode's editorial solutions, which are far too dense and wordy to be of much use to leetcode students. Only the most focused and motivated would be able to parse through that crap. Thanks for sharing this with the world, you're making my life a lot better.

  • @dhruvsolanki4473
    @dhruvsolanki4473 18 дней назад +2

    We can use array to hold frequency as string is only consist of uppercase english letters.

  • @shriyanshkhandelwal3988
    @shriyanshkhandelwal3988 2 года назад +1

    I love the way you explain and even teaching DSA with python. Loved it TY

  • @celialiang1485
    @celialiang1485 3 года назад +1

    Very helpful video. I learned how to build up the idea from scratch. It's needed especially when you explain to the interviewer in a tech screening. Thanks a lot!

  • @alcatraz5161
    @alcatraz5161 10 месяцев назад +1

    the best part about this solution is how it is character agnostic
    a problem like this you'd want to think about chaining bits and pieces of character sub-strings as long as the gap size between them is under current value of k but this removes the necessity for that
    well solved and thank you :)

    • @manasisingh294
      @manasisingh294 4 месяца назад +1

      general solution is always likely infinite times better.

  • @VivekMishra-hd7mg
    @VivekMishra-hd7mg 3 года назад +3

    He gave the best explanation. really good man!! This channel will grow a lot in future.

  • @saravalls26
    @saravalls26 7 месяцев назад

    The max(count.values()) amazed me, I had written an extra function to look for that. You're the GOAT of this man, thanks a lot for your hard work ♥

  • @SkuZY22
    @SkuZY22 21 день назад

    Amazing video! Note that you can simply do: res = r - l + 1 without using max, and still get the correct answer (with fewer operations).

  • @rohandevaki4349
    @rohandevaki4349 2 года назад +34

    how do you approach to tough problems like this?and how can we improve our thinking and optimise the code? i am not even able to think of a bruteforce approach for this, it is that hard.

    • @cschandragiri
      @cschandragiri 2 года назад +4

      this one is hard

    • @yuurishibuya4797
      @yuurishibuya4797 Год назад +8

      Keep solving, that will create a hash map in your brain, problem pattern - solution. Then when u look at different problems, you will know how to solve in O(k) or O(log(n))

    • @MrjavoiThe
      @MrjavoiThe Год назад +2

      @@yuurishibuya4797 how can you solve more, if you can’t solve anything and trying to solve by your own confuses you more and brings the habit of doing horrible solutions.

    • @MrjavoiThe
      @MrjavoiThe Год назад +1

      @@yuurishibuya4797 isn’t better to understand the different solutions and patterns and then try to apply those yourself on similar problems?

    • @udaykulkarni5639
      @udaykulkarni5639 Год назад +18

      Practice and consistency
      Dont worry about the optimization. Optimization usually requires a trick that you learn when you keep solving problems. Try to come up with brute force solution within say 30 mins. If you cant. Just look at solution. Code it yourself. Understand it thoroughly. Make a note of it. Come back to same problem after 1 week. Can you still code it? This time you looked at problem differently. And same thing happens when you keep doing it.
      But but but. It takes lot of patience and discipline. You will feel like giving up and thinking its not for you. But remember - IT IS DIFFICULT AND IT TAKES TIME. YOU WILL MAKE IT CHAMP!! GOOD LUCK

  • @HumbleShoeLace
    @HumbleShoeLace Месяц назад

    How did I not find this channel earlier. youe explainations are amazing. it becomes so easy to understand.

  • @anaykulkarni8956
    @anaykulkarni8956 26 дней назад

    Best explanation I have seen so far on this topic! great job

  • @RanjuRao
    @RanjuRao 3 года назад +15

    Awesome as always !! Love to watch your videos ! Please do the system design videos as you're very good in explaining concepts :)

  • @pragyasingh5880
    @pragyasingh5880 2 года назад +4

    Damn this is the best explanation of this problem so far. Thanks for this awesome video!

    • @NeetCode
      @NeetCode  2 года назад

      Happy it's helpful! 🙂

  • @gokulnaathbaskar9808
    @gokulnaathbaskar9808 2 года назад +1

    The first approach was very easy to understand, thank you so much

  • @seanwayland
    @seanwayland 7 месяцев назад

    An amazing channel. Comments in the code solution would make it easier to understand and if I was an interviewer more likely to give the candidate a job.

  • @avipatel1534
    @avipatel1534 2 года назад +5

    Just one minor thing, you could replace the inner while loop with an if statement because you would be iterating a max of 1 time in that while loop

    • @mihirmodi1936
      @mihirmodi1936 Год назад +1

      Yes Right!! But It will be good if you explain it that why we don't need loop. (Try with this example s="AABCDAAAAMN" k=2) Here, We need to execute loop multiple time, but still if will give correct ans.

    • @InfinityOver0
      @InfinityOver0 2 месяца назад

      @@mihirmodi1936 Why? Please tell me.

    • @OmniscientDreamer
      @OmniscientDreamer Месяц назад

      @@InfinityOver0 It works with if, because we anyways will iterate to the next j pointer and check the condition for k

  • @EE12345
    @EE12345 11 месяцев назад +2

    Hey, do you guys think the O(26N) to O(N) optimization would actually make the difference in an interview? To me the whole idea that we can ignore maxf while shrinking the window is hard to understand and especially to deduce in an interview setting.

  • @kaos092
    @kaos092 8 месяцев назад +1

    This should be a leetcode hard. This is way harder than the water trap question.

  • @nimeshpareek953
    @nimeshpareek953 Год назад

    Man you just explained it this easy like I watched the video and write pseudocode on pen and paper and coded it and in single go submitted

  • @sandeshsrinivas4177
    @sandeshsrinivas4177 2 года назад

    This approach is far more superior than any other solution.

  • @discover_afrika
    @discover_afrika 5 месяцев назад

    Nice video, because moving the left pointer once would always make the condition "(r - l + 1) - max(count.values()) > k" True, you can replace the inner while loop with an "if" statement to better show that the algorithm is linear in time.

  • @phuongdh
    @phuongdh 3 года назад

    Finally someone who explains why we didn't need to decrease the max_frequency, other videos offer no explanation.

  • @Han-ve8uh
    @Han-ve8uh 2 месяца назад

    3:38 "We want all characters in window to match most common character".
    I struggled to believe this greedy strategy gives the best answer, and came up with a counter-example BCDBCDBAZA and k=1.
    In this case B is most common with 3 but using the k on A gives the best answer of 3 instead of 2 if used for B.
    Then I realized this example would be an impossible window if we followed the sliding window technique which would have long ago moved the left pointer.
    Lesson is to trace through an example in more detail by actually following through an iteration strategy, instead of jumping into the middle of some loop and checking scenarios, and the fact that there's a chicken-egg issue.
    Before coming up with solution, we need test cases, but the test cases are influenced by iteration strategy.
    Here's a case of using impossible test cases that makes finding solution impossible. If someone does not know sliding window, it would be very difficult to even know that BCDBCDBAZA is a bad example.
    How do others deal with this "wrong test case leading to false negative strategy" situation? Are there better answers than "Memorize every solution strategy, throw all and see what sticks?"

  • @fuadmohamoud2856
    @fuadmohamoud2856 3 месяца назад

    This problem is tricky to get right. Thanks for the video Neetcode!

  • @chetan788
    @chetan788 2 года назад +3

    Did you figure out all solutions on your own? I could only figure out only 2 or 3 out of 14 mediums I have done so far.
    What am I missing? How could I improve?

  • @yadhunandhanr7590
    @yadhunandhanr7590 4 месяца назад

    I was keeping track of the frequent character and based on that was updating an internal K value along with max frequency. But that made the code too complex, had to handle lot of cases. But when I saw your solution, I couldn't believe how simple you made it.

  • @daddac
    @daddac 3 года назад

    i cant figure out why max frequency works even the char might out of sliding window... until i watch this video. awesome indeed. thanks for posting such concise and clean video!!

  • @kvtys
    @kvtys 3 месяца назад

    The maxf is so tricky - but its one of the tricks I can genuinely appreciate. It requires us to really understand the fundamental logic of this problem for it to even be a thought in our heads

  • @onlylikenerd
    @onlylikenerd 2 года назад +1

    I spent over 3 hours solving this on my own. My solution was 80 lines of code. Here I am looking at your solution mind blown.... lol. Boy did I over complicate things.

  • @niteshmgs7569
    @niteshmgs7569 3 года назад +1

    Great job mate, very good explanation... keep working, your channel will definitely grow. It's the best

  • @abhirupchakravarty8508
    @abhirupchakravarty8508 Год назад

    if we store the curr max f and the char that gives it, when we add a new char in the window, we can increment and check its frequency, and if that's higher than max f, we can update both the most f char and the max f to the newly added char (similarly, in decrements, but as NC pointed out, that's not required)

  • @yuanpengli5704
    @yuanpengli5704 2 года назад +1

    Absolutely love your teaching style

  • @pallisaiprasad618
    @pallisaiprasad618 2 года назад +4

    Thanks for the explanation.
    Quick question: How is the time complexity O(n)?
    The while loop inside, is that considered constant time complexity?

    • @EquinoXReZ
      @EquinoXReZ Год назад

      Did you figure this out? Also confused about this

    • @mgik-or-something
      @mgik-or-something 9 месяцев назад

      Hey, I might be a bit late but if it helps. Basically time complexity of O(n) means that with the number of operands increasing, time complexity will increase linearly. So basically if we had like 5 operands, the time complexity would be x operations, but then if we had 6 operands, the time complexity would be x + (some constant) and so on and so on. The same applies to the while loop, where for the given amount of operands the while loop will perform the same amount of iterations corresponding to the number of operands. And in the end we would still have the O(n) time complexity because of that.

  • @souravsanyal2612
    @souravsanyal2612 Месяц назад

    For this problem, the solution suggests, use two-pointer method to make sure the substring is valid.
    The condition for that is (right-left+1) - max(count.values()) > k.
    And the examples provided only show two characters.
    What happens if there are more than 2 characters ? If we even replace the maximum occurring character, what is the guarantee the remaining characters are all the same? Attaching the code which is passing .
    class Solution:
    def characterReplacement(self, s: str, k: int) -> int:
    count = {}
    left = 0
    res = 0
    for right in range(len(s)):
    count[s[right]] = count.get(s[right],0)+1
    while right-left+1 - max(count.values()) > k:
    count[s[left]] -= 1
    left += 1
    res = max(res,right-left+1)
    return res

  • @u2bacount
    @u2bacount 2 года назад +1

    For anyone still struggling to grasp the final optimization with maxf, I think the following can help understand.
    The algorithm satisfies the following 2 invariants:
    1. line 10 updates maxf ito be the maximum frequency observed in the substring between l and r up until that point.
    2. at line 16, the number of characters between l and r (which r - l + 1) is never greater than the longest legal substring seen up until that point

    • @u2bacount
      @u2bacount 2 года назад +3

      To make this point clearer, there is no point in updating res at line 16 as the result is completely determined by maxf.
      Here is an additional optimization which instead of updating res on every iteration of the while loop just derives the result from maxf
      class Solution:
      def characterReplacement(self, s: str, k: int) -> int:
      count = {}
      l = 0
      maxf = 0
      for r in range(len(s)):
      count[s[r]] = 1 + count.get(s[r], 0)
      maxf = max(maxf, count[s[r]])
      while (r - l + 1) - maxf > k:
      count[s[l]] -= 1
      l += 1
      return min(maxf + k, len(s))

    • @ErZi-uo7fm
      @ErZi-uo7fm 3 месяца назад

      Thanks! This finally clicked after reading a bunch of different responses.

  • @anishr8017
    @anishr8017 3 года назад +2

    You're doing an awesome job!! Thank you! Keep up the work, you're helping a lot of us!

  • @quirkyquester
    @quirkyquester 5 месяцев назад

    bruhh,, you make this difficult questions seems so easy, thank you!

  • @biswajitmahalik4134
    @biswajitmahalik4134 Год назад

    thanks a lot, man! I saw the codes n wasn't able to figure out why aren't they downgrading max freq...n you explained it so nicely .

  • @sukinkumar7042
    @sukinkumar7042 2 года назад +1

    Incredible. Especially the piece of math from 13:53! Too good :)

  • @dgvj
    @dgvj 2 года назад

    To comeup with the condition was really difficult.Thanks a ton for the awesome explanation.

  • @pabloarkadiuszkalemba7415
    @pabloarkadiuszkalemba7415 2 года назад +1

    Another way of doing it is mantaining a max heap to get the max value, so it would be O(log(26))

  • @ivandrofly
    @ivandrofly Месяц назад

    Sliding window is one of the most essential algorithm someone can learn!

  • @bens9962
    @bens9962 2 года назад

    Hey, just want to say that you're the best :D Your visualization makes a lot of difference. Thank you!

  • @kuoyulu6714
    @kuoyulu6714 Год назад +1

    another great day, with another great neetcode video

  • @abhiramchowdary5121
    @abhiramchowdary5121 2 года назад +1

    To help others who might find difficulty understanding why Maxf is not decrementing .the key is maxf is cannot be greater than maxf or count[s[r]] in the existing window .the rough explanation is unless the s[r] & s[r+1] are equal, Maxf will not increase ,if both are different you will get wrong res for that window but still it happens in previous window so we can ignore , if you might worry s[r] and s[r+1] are different but we still get Maxf same no problem since it will also give the same count. You might also worry in another case s[r] and s[r+1] are different and s[r] and s[r+2] are same ,
    then Maxf might increase but thanks to decrement in while loop count[s[r]] will decrease and maxf will still be same.

    • @DED_Search
      @DED_Search Год назад +1

      I honestly do not understand what you talked about.

  • @nguyenquoctran8423
    @nguyenquoctran8423 Год назад +3

    If we enter the while condition and move our left part, we will decrease count[s[left]]. What if s[left] is the most frequent char ? Shouldn't maxf be updated too ? Because in the iteration after that, we would compare the old maxf (which counts some char that have been cut off the window) with count[right] and assign the max of that to maxf. Woulnd't that overestimate maxf ?

  • @wagle04
    @wagle04 10 месяцев назад

    18:00 you have to use while there. if does not work when k=0. found it out the hard way lol

  • @jeremydavidson7559
    @jeremydavidson7559 3 месяца назад

    Not only is the if statement sufficient, it is more efficient. If you used the while loop, you would be incrementing left, but keeping right static and therefore never finding a window longer than the longest already found. However if you did the if statement you're never doing unnecessary operations on smaller sized windows, and will reach the end condition of the for loop faster.

  • @NickolasCavagnaro
    @NickolasCavagnaro 2 года назад

    We can further optimize from O(n * m) where m is the number of characters to just O(n) where n is the size of the input array.
    Once we have a window of size k + majority count, we don't need to update the majority count as a result of shrinking the window as long as we only increase the result when we have encountered a higher majority count than before.
    And before we reach a window size of k + majority count, the majority count will be correct because it will not include characters from before the window. But once we do have a window of size k + majority count, we can always keep the window size the same, only increasing it if we encounter a higher majority count than before, and only updating the result when we have updated the window size.
    This way, we don't need to go through the counts of each character, we only need to update the majority count to be the highest majority count we have seen in a window so far.

  • @emmisae-ueng8976
    @emmisae-ueng8976 3 года назад

    The explanation is really good, easy to follow and understand.

  • @justforfun4680
    @justforfun4680 8 месяцев назад

    How I understood maxf is that: Only if you can get a bigger maxf (under constraint of window_length - maxf

  • @tomonkysinatree
    @tomonkysinatree 9 месяцев назад

    I think the easiest way to think about the maxf optimization is this: the res (length) is not updated unless the condition len - maxf

  • @adityaojha2701
    @adityaojha2701 3 года назад +2

    Thanks for such a nice and clear explanation!!

  • @m_jdm357
    @m_jdm357 6 месяцев назад

    From what I saw it can't get simpler than the solution NeetCode provided. Thank you for the simplest solution. No, no, no this thing is not solvable if you don't know the solution.

  • @free-palestine000
    @free-palestine000 2 года назад +1

    Wow this took me so long to understand, thank you

  • @ujjvalsharma5055
    @ujjvalsharma5055 3 года назад

    Wonderful solution!! Keep making videos! You will definitely get more subscribers!!

  • @aniketbhanderi5927
    @aniketbhanderi5927 Год назад +1

    We can use int[] array of size 26 to reduce the space complexity hashmap from O(n) to O(26)=O(1). Surprisingly, it takes much less time too.

    • @rajkumarkhatri8119
      @rajkumarkhatri8119 Год назад

      Because we can have no more than 26 keys in the hashmap, space complexity of hashmap (in this case) is O(26)~O(1).

  • @gregoryvan9474
    @gregoryvan9474 2 года назад

    great video! I think it makes sense though to use an "if" condition instead of the "while" loop because you are checking that condition twice unnecessarily. once you move that left pointer over once, you will always break while loop because (r - l - 1) - maxf > k will return false.
    def characterReplacement(self, s, k):
    count = {}
    if k+1 >= len(s):
    return len(s)
    l = 0
    r = 0
    maxcount = 0
    while r < len(s):
    count[s[r]] = 1 + count.get(s[r], 0)
    maxcount = max(maxcount, count[s[r]])
    if (r-l+1) - maxcount > k:
    count[s[l]] -= 1
    l += 1
    r += 1
    return r-l

  • @saitrinathdubba
    @saitrinathdubba 2 месяца назад

    This is so beautifully explained !! and very clever problem (Y)

  • @protyaybanerjee5051
    @protyaybanerjee5051 2 года назад

    Thank you for creating this channel.

  • @downtownsocialite1206
    @downtownsocialite1206 2 года назад

    Absolutely clean explanation, good work!

  • @arthurasanaliev
    @arthurasanaliev Год назад

    wow, such an elegant solution, thanks for explanation!

  • @bombdotcomist
    @bombdotcomist Год назад +1

    I don't understand at all why the "res = max(res, r-l +1)" happens on every iteration. If we have the example AABAB where k=1, A works, AA works, AAB works, AABA works, but AABAB doesn't work, so we reduce the window and now the A on the left disappears and we have _ABAB. According to this code, we update the result to now be 4, but "ABAB" is not a valid window. I'm so confused...

  • @smartlab5173
    @smartlab5173 2 года назад

    You are the Greatest of the Great! Champion! Honestly, words are not enough to thank you. Your clarity of explanation is Mind Blowing. God bless you! :-)

  • @legendsmnd
    @legendsmnd Год назад

    Correct me if I'm wrong, you do not need to decrement the max frequency variable because that won't affect the result. It is an overestimation and the following characters will produce the same result until there is an update in the max frequency, which increases the result.

  • @leonardmensah6781
    @leonardmensah6781 2 года назад

    This is what we call a G.O.A.T explanation !!!!!!

  • @ToastFrench24
    @ToastFrench24 2 года назад +1

    How is the first solution O(26n)? Summing up n + (n-1) + (n-2) + ... should still be O(n^2)?

  • @StLegend950111
    @StLegend950111 2 года назад

    Dude, you are awesome. Thanks for your clean explanation.

  • @zeroblade8315
    @zeroblade8315 3 года назад +1

    why does both an if and a while loop work for the solution, for shifting the left pointer?

  • @algosavage7057
    @algosavage7057 3 года назад

    Thanks a lot for your explanation. I was so curious about solving in more efficient than O(26*n).

  • @apdr8783
    @apdr8783 3 года назад +4

    once you start shrinking the window, the max count is no longer the max within the window, but a somewhat global max in the whole substring that you have scanned so far, this count is all screwed up, isn't it? I still don't get it.

    • @tb8588
      @tb8588 3 года назад

      hmm yea that is the part that I don't understand either lol, how does maxf stay constant when we do the check in the while loop lmao

    • @Jessie-vz9wd
      @Jessie-vz9wd 2 года назад

      Maybe since you are only shifting the left pointer while the right pointer is fixed, the new max count will not exceed the previous one, and the res will not change as well when you shrink the sliding window. I wrote a similar solution which might make it clearer :)
      class Solution:
      def characterReplacement(self, s: str, k: int) -> int:
      l = 0
      res = 0
      cnt = {}
      maxf = 0
      for r in range(len(s)):
      cnt[s[r]] = 1 + cnt.get(s[r], 0)
      maxf = max(maxf, cnt[s[r]])
      if r-l+1 - maxf k:
      cnt[s[l]] -= 1
      l += 1
      return res

  • @deepakheerakari807
    @deepakheerakari807 2 года назад

    class Solution:
    def characterReplacement(self, s: str, k: int) -> int:
    l=0
    r=0
    maxi=0
    ans=0
    temp=0
    n=len(s)
    hm={}
    while(rk:
    #shrinking the window and decreasing most frequent character
    hm[s[l]]-=1
    l+=1

    ans=max(ans,r-l+1)
    r+=1
    return ans

  • @prunesalad
    @prunesalad 2 года назад

    thanks for going deep on the O(1n) case, very interesting

  • @jackieli1724
    @jackieli1724 Год назад

    Needless to update maxf when l increase and (r - l + 1) decrease, ((r - l + 1) - maxf) will stay the same if s[l] is the character of max frequency, ((r - l + 1) - maxf) will decrease when s[l] is not character of max frequency or just one of characters of max frequency.

  • @andreytamelo1183
    @andreytamelo1183 2 года назад

    Thanks!

  • @AniketSingh-hr8mi
    @AniketSingh-hr8mi 2 года назад +1

    how do you come up with solutions. i try for hours and then maybe get a naive solution which probably gets time limit exceed error

  • @licokr
    @licokr 8 месяцев назад

    Plus, there's a room you can improve the logic a little bit. instead of max(count.values()), you may define a variable, let's say 'currentMax' then you can replace max(count.value()) with max(count[s[r]], currentMax). Then, it doesn't need to iterate all the values, instead it compares only two values. Thanks for the video neetcode 👍