-
Couldn't load subscription status.
- Fork 39
Earth - Christina Minh #23
New issue
Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.
By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.
Already on GitHub? Sign in to your account
base: master
Are you sure you want to change the base?
Changes from all commits
File filter
Filter by extension
Conversations
Jump to
Diff view
Diff view
There are no files selected for viewing
| Original file line number | Diff line number | Diff line change |
|---|---|---|
| @@ -1,19 +1,52 @@ | ||
|
|
||
| # This method will return an array of arrays. | ||
| # Each subarray will have strings which are anagrams of each other | ||
| # Time Complexity: ? | ||
| # Space Complexity: ? | ||
|
|
||
| # Time Complexity: O(n * m * log m) | ||
| # where n is the number of strings in the array to iterate over, | ||
| # and m is the number of characters to sort in a string (m log m), | ||
| # Space Complexity: O(n * m) from sorting characters (m) and returning a new array (n) | ||
| # where n is the number of strings in the array | ||
| # and m is the number of characters in a string | ||
| def grouped_anagrams(strings) | ||
| raise NotImplementedError, "Method hasn't been implemented yet!" | ||
| sorted_strings_hash = Hash.new([]) | ||
|
|
||
| strings.each do |word| | ||
| word.chars.sort.join | ||
|
|
||
| if sorted_strings_hash[word] | ||
| sorted_strings_hash[word].push(word) | ||
| end | ||
| end | ||
|
|
||
| puts sorted_strings_hash.values | ||
| return sorted_strings_hash.values | ||
| end | ||
|
|
||
| # This method will return the k most common elements | ||
| # in the case of a tie it will select the first occuring element. | ||
| # Time Complexity: ? | ||
| # Space Complexity: ? | ||
| # Time Complexity: O(n + m log m) where n is the number of elements in the list and | ||
| # and m is the number of unique elements to sort (m log m) | ||
| # Space Complexity: O(m * k) where m is the number of unique elements to put in hash | ||
| # and k is the number of elements to return | ||
| def top_k_frequent_elements(list, k) | ||
|
Comment on lines
+27
to
31
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 👍 I would say this is O(n log n) since k < n, and you're potentially sorting a hash of n elements. For space complexity since k < n, and you make a hash of potentially n elements this is O(n). |
||
| raise NotImplementedError, "Method hasn't been implemented yet!" | ||
| return [] if list.empty? | ||
|
|
||
| frequency_hash = Hash.new(0) | ||
|
|
||
| list.each do |num| | ||
| frequency_hash[num] += 1 | ||
| end | ||
|
|
||
| sorted_frequency_hash = frequency_hash.sort_by { |num, frequency| -frequency } | ||
|
|
||
| k_frequent_elements = [] | ||
|
|
||
| k.times do |i| | ||
| num_frequency_pair = sorted_frequency_hash[i] | ||
| k_frequent_elements << num_frequency_pair[0] | ||
| end | ||
|
|
||
| return k_frequent_elements | ||
| end | ||
|
|
||
|
|
||
|
|
@@ -22,8 +55,54 @@ def top_k_frequent_elements(list, k) | |
| # Each element can either be a ".", or a digit 1-9 | ||
| # The same digit cannot appear twice or more in the same | ||
| # row, column or 3x3 subgrid | ||
| # Time Complexity: ? | ||
| # Space Complexity: ? | ||
| # Time Complexity: O(1) constant time to iterate over table that is always 9x9 | ||
| # Space Complexity: O(1) | ||
| def valid_sudoku(table) | ||
|
Comment on lines
+58
to
60
There was a problem hiding this comment. Choose a reason for hiding this commentThe reason will be displayed to describe this comment to others. Learn more. 👍 Nicely done. |
||
| raise NotImplementedError, "Method hasn't been implemented yet!" | ||
| end | ||
| seen_in_rows = [] | ||
| seen_in_columns = [] | ||
|
|
||
| # array index represents the row/column number | ||
| 10.times do |i| | ||
| seen_in_rows[i] = {} | ||
| seen_in_columns[i] = {} | ||
| end | ||
| # Example seen_in_rows = [ | ||
| # {"5"=>true, "3"=>true, "7"=>true}, row 0 | ||
| # {"6"=>true, "1"=>true, "9"=>true, "5"=>true}, row 1 | ||
| # {"9"=>true, "8"=>true, "6"=>true}, row 2 | ||
| # ... | ||
| # ] | ||
|
|
||
| seen_in_subgrids = | ||
| [ | ||
| [{}, {}, {}], | ||
| [{}, {}, {}], | ||
| [{}, {}, {}], | ||
| ] | ||
|
|
||
| num_rows = table.length | ||
| num_columns = table[0].length | ||
|
|
||
| num_rows.times do |row| | ||
| num_columns.times do |col| | ||
| current_cell_value = table[row][col] | ||
|
|
||
| next if current_cell_value == "." | ||
|
|
||
| # find position of hash in seen_in_subgrids | ||
| subgrid_col = col / 3 | ||
| subgrid_row = row / 3 | ||
|
|
||
| if seen_in_rows[row][current_cell_value] || seen_in_columns[col][current_cell_value] || seen_in_subgrids[subgrid_row][subgrid_col][current_cell_value] | ||
| return false | ||
| end | ||
|
|
||
| # Add value to all look-up hashes | ||
| seen_in_rows[row][current_cell_value] = true | ||
| seen_in_columns[col][current_cell_value] = true | ||
| seen_in_subgrids[subgrid_row][subgrid_col][current_cell_value] = true | ||
| end | ||
| end | ||
|
|
||
| return true | ||
| end | ||
There was a problem hiding this comment.
Choose a reason for hiding this comment
The reason will be displayed to describe this comment to others. Learn more.
👍