Rewriting 15-Year-Old Code

Did you ever have to maintain a fifteen year old application? Dead code everywhere, database queries in between HTML tags and some pages still in PHP 3. This strategy-level presentation will lead you through a progressive rewrite from very old legacy to the latest shiny version of PHP. Learn how to automate legacy testing, how to seamlessly jump between the old and new parts, and how to overcome other challenges that arise from dealing with legacy.

Transcript

1. @afilina
   Rewriting 15-Year-Old Code
   IPC, Berlin - May 30, 2017
   

   View Slide

2. Anna Filina
   • whoami():\LegacyFixer
   • Project rescue expert
   • Dev, trainer, speaker
   

   View Slide

3. You inherited a

   10-year-old codebase
   

   View Slide

4. You inherited a

   15-year-old codebase
   

   View Slide

5. You inherited a

   17-year-old codebase
   

   View Slide

6. You inherited a

   17-year-old codebase
   Code doesn't age like wine.
   

   View Slide

7. The Web Was 600px Wide
   

   View Slide

8. Feeling Old Yet?
   • Y2K bug.
   • Internet Explorer 5.
   • ICQ was starting to be cool.
   • Rounded corners = status symbol.
   

   View Slide

9. Code smells
   

   View Slide

10. Mixed Concerns
    mysql_query('...');
    $total = 0;
    foreach ($products as $p) {
    $total += $p['qty'] * $p['price'];
    }
    ?>
    Total: = $total ?>
    

    View Slide

11. Poorly Named Variables
    $a = //...
    $array = //...
    $item3 = //...
    

    View Slide

12. Global Functions & Constants
    include("functions.php");
    MyClass::myMethod(); // namespaced function?
    myMethod(DEV_MODE);
    

    View Slide

13. Inexplicable Conditions
    if ($order_id > 20117) {
    // use this sql
    } else {
    // use that sql
    }
    // weakness = time to refactor
    

    View Slide

14. Long Methods
    public function importCsv($path, $googleApiKey, $databaseDsn, $databaseUser, $databasePassword)
    {
    // Convert CSV to array of conferences
    $lines = file($path);
    $csv = array_map('str_getcsv', $lines);
    $conferences = [];
    foreach ($csv as $line) {
    $conference = new Conference();
    $conference->name = $line[0];
    $conference->location = $line[1];
    $conferences[] = $conference;
    // Get coordinates for location
    $location = urlencode($conference->location);
    $url = 'https://maps.googleapis.com/maps/api/geocode/json?address='.$location.'&key='.$googleApiKey;
    $curl = curl_init();
    curl_setopt($curl, CURLOPT_URL, $url);
    curl_setopt($curl, CURLOPT_RETURNTRANSFER, 1);
    $response = curl_exec($curl);
    curl_close($curl);
    $json = json_decode($response);
    if (count($json->results) == 0) {
    continue;
    }
    $latitude = $json->results[0]->geometry->location->lat;
    $longitude = $json->results[0]->geometry->location->lng;
    $coordinates = $latitude.','.$longitude;
    $conference->coordinates = $coordinates;
    // Save conference to database
    $pdo = new PDO($databaseDsn, $databaseUser, $databasePassword);
    $statement = $pdo->prepare('REPLACE INTO conference (name, location, coordinates) VALUES (?, ?, ?)');
    $statement->execute([
    $conference->name,
    $conference->location,
    $conference->coordinates,
    ]);
    

    View Slide

15. Average Legacy Code File
    • 3000-6000 lines of code.
    • Half of it is commented "in case we need
    it later".
    • Method length of 800 lines.
    • Abuse of helper classes.
    • Sometimes no classes at all.
    

    View Slide

16. Codebase
    • 5,000 classes.
    • 20,000 methods.
    • 1,500,000 lines of code.
    

    View Slide

17. Before you code
    

    View Slide

18. Strategy
    • Make a strategy based on
    constraints.
    • Full rewrite vs progressive:
    ◦By class.
    ◦By feature/module.
    ◦By HTTP call.
    • How to run code side-by-side:
    ◦Session sharing.
    ◦mod_rewrite.
    

    View Slide

19. Data Can Be Lost, Stuff Can Break
    • Backup: test restore.
    • Nullify sensitive data.
    • Staging: simulate deployments/
    upgrades/batch processes.
    • Automate tests before code
    changes.
    • Make a risk assessment:
    ◦ Don't be too optimistic.
    ◦ Account for side-effects.
    

    View Slide

20. Rewrite example
    When it goes wrong.
    

    View Slide

21. PHP 3 to PHP 5.6
    • HTML + PHP + SQL in same file.
    • Includes all over the place.
    • IFs that concatenate SQL.
    • Previous rewrite attempt:
    ◦ Failed, made things worse.
    ◦ Folders of dead code.
    ◦ Classes with static functions (no instances).
    

    View Slide

22. Solution
    • Rewrite complex forms in Symfony:
    ◦ mod_rewrite for concerned pages.
    • Rewrite biggest module as OOP:
    ◦ Design extraction.
    ◦ Flexible architecture.
    ◦ Automated tests.
    

    View Slide

23. Design extraction
    

    View Slide

24. Avoid Code Bias
    • Old code → design docs.
    • Validate design docs:
    ◦Clarify business rules.
    • Improve design:
    ◦Reduce technical debt.
    ◦More flexible.
    • Design docs → new code.
    

    View Slide

25. Fixing bugs
    

    View Slide

26. Code Duplication
    • Sometimes, the bug is
    repeated 80+ times.
    • Remove duplications
    ASAP.
    

    View Slide

27. Fix Long Methods
    • Extract broken part into its own method.
    • Write unit tests for it.
    • Fix it.
    • Call it from the mega-method.
    

    View Slide

28. Spot Logical Groups
    Code block
    Comment
    Code block
    Comment
    Convert CSV to array of conferences.
    Get coordinates for conference location.
    Save conference to database.
    Code block
    Comment
    

    View Slide

29. Extract & Test
    Code block
    Comment
    Code block
    Method
    Code block
    Comment
    Call method
    

    View Slide

30. Name
    Code block
    Method
    Code block
    Method
    Call method
    Call method
    Code block
    Method
    Call method
    getConferencesFromCsv
    getLocationCoordinates
    saveConference
    

    View Slide

31. Rewrite example #2
    

    View Slide

32. ASP Classic to PHP 5.6
    • 15+ spaghetti and hacks.
    • Language no longer supported.
    • Huge ERP with lots of code.
    

    View Slide

33. Solution
    • Rewrite page by page to framework.
    • mod_rewrite for concerned pages.
    • DB session adapter in both apps.
    • Page in any language = HTTP request.
    ◦ Guzzle tests FTW!
    

    View Slide

34. Guzzle Tests
    

    View Slide

35. On Testing
    Before you code
    You better test
    So that on weekends
    You may rest.
    -- Anna Filina
    

    View Slide

36. Testing Static Methods
    $mock = Mockery::mock('alias:SomeClass');
    $mock->shouldReceive('staticMethod')->andReturn(42);
    

    View Slide

37. Stuck?
    

    View Slide

38. Try Something New
    • Bounce ideas.
    ◦ New people to avoid tunnel vision.
    • Has this been done before?
    • Can I try another approach?
    

    View Slide

39. Takeaways
    • Make a strategy.
    • You touch it, you refactor it.
    • Use known tools & methodologies.
    • Get inspiration from others.
    • Refactoring gets easier.
    • Every problem has a solution.
    

    View Slide

40. @afilina afilina.com
    joind.in
    

    View Slide