franckyu/batting-averages-cli

Batting Average Command line app

Batting Averages Backend Exercise

Batting average is simple and a common way to measure batter’s performance. Create an app that will ingest a raw CSV file with player statistics and provide will provide player rankings based on their batting performance.

Input

The application should take an input in form of a CSV file. The file will be comma separated CSV with headers. The headers that interest you are: “playerID”, “yearID”, “stint”, teamID”, “AB”, and “H”.

The application should also accept filter options:

• Year
• Team name
• Year and Team name

When filter are present, the output should include only players that match the filter, sorted according to their batting average.

Expected output

Batting Average is calculated as: BA = H/AB (Hits / At Bats).

If the player has more stints in the season, calculate batting average for the whole season (across all stints), team names are comma separated in that case. Format the batting average to 3 decimals.

Write a standalone command line app

The output should be like a following table:

+----------+--------+--------------+-----------------+
| playerID | yearId | Team name(s) | Batting Average |
+----------+--------+--------------+-----------------+
| ...                                                |
+----------------------------------------------------+

CSV files

The input CSV file is in data/Batting.csv. This file includes "teamID", use the file data/Teams.csv to map "teamID" to a team's real name.

---

Exercice Implementation

A. Installation

1. Use RVM or rbenv
2. clone this repository and cd into it, then bundle install

The dependencies are:

• Thor for the main command entry
• tty-spinner for the UI when the CSV is loading. From the excelent TTY toolkit suite.
• tty-table to format the ranking results in ASCII tables
• Paint to color the blue prompt
• SortedArrayBinary that allows for fast sorting on items insertion in the rank.

From the Standard Library:

Set for fast filtering by set intersection and duplicate cleanup.

B. Running the CLI app

This app uses Thor for the main command. Just execute thor display:batting_averages_ranking to launch it.

C. UI

Once launched, you'll play with an interactive REPL which on first time loads and process the CSV data (100k+ lines).

The first result will be the all time top 30 batting average rankings

Just below it you have a blue prompt |> where you enter specific commands to play with the ranking table

D. Commands

Type h or help (case insensitive) to display available commands. They are self-explanatory:

1. all or ENTER -> displays global ranking
2. in {4_digits_year} -> displays ranking for this year
3. of {team_name} -> displays ranking within this team from its inception
4. of {team_name} in {4_digits_year} -> combination of 2 and 3
5. list years
6. list teams
7. n or next -> go to next page
8. p or previous -> back to previous page
9. q or Ctrl+D -> exit

E. Pagination

The results are paginated by batches of 30 lines from top ranked player per year.

Type n or next to descend the ranking table. Or p or previous to go back to upper ranks.

F. Test

Quite simple, just run rake or rake test to launch the test

The test file skips the CSV parsing and jumps directly to use parsed data. A global test matrix is provided on top for better understandability.

G. Internal architecture

The app is separated in 3 layers

1. The Thor command that launches the app, in cli.thor.
2. UI layer, implemented in lib/batting_averages_ui.rb. This layer handles the output presentation, the textual interactivity and the REPL loop
3. Engine layer, implemented in lib/batting_averages_engine.rb. This layer is responsible of the CSV parsing and ranking calculations

Strategies

• Caching of the global ranking because this is heavy to process (100k+ lines and we cannot leverage indexes)
• Ahead of time indexing by year and by team. This speeds up the parsing when filtering is asked by the user (no need of cache)
• Usage of Set: provides a time complexity advantage when filtering is asked by the user (~ set intersection) + automatic data cleaning (removal of duplicate rows)
• Usage of SortedArrayBinary that is faster than native Array for sorting items.
• For readability in the UI, the tables are presented in pages of 30 by 30, which is reset to page 1 (top ranks) whenever the filtering criteria changes