feat(scripts): add sportstime-parser data pipeline

Complete Python package for scraping, normalizing, and uploading sports schedule data to CloudKit. Includes: - Multi-source scrapers for NBA, MLB, NFL, NHL, MLS, WNBA, NWSL - Canonical ID system for teams, stadiums, and games - Fuzzy matching with manual alias support - CloudKit uploader with batch operations and deduplication - Comprehensive test suite with fixtures - WNBA abbreviation aliases for improved team resolution - Alias validation script to detect orphan references All 5 phases of data remediation plan completed: - Phase 1: Alias fixes (team/stadium alias additions) - Phase 2: NHL stadium coordinate fixes - Phase 3: Re-scrape validation - Phase 4: iOS bundle update - Phase 5: Code quality improvements (WNBA aliases) Co-Authored-By: Claude Opus 4.5 <noreply@anthropic.com>
2026-01-20 18:56:25 -06:00
parent ac78042a7e
commit 52d445bca4
76 changed files with 25065 additions and 0 deletions
--- a/sportstime_parser/tests/test_scrapers/test_mlb.py
+++ b/sportstime_parser/tests/test_scrapers/test_mlb.py
@@ -0,0 +1,257 @@
+"""Tests for MLB scraper."""
+
+from datetime import datetime
+from unittest.mock import patch
+
+import pytest
+
+from sportstime_parser.scrapers.mlb import MLBScraper, create_mlb_scraper
+from sportstime_parser.scrapers.base import RawGameData
+from sportstime_parser.tests.fixtures import (
+    load_json_fixture,
+    MLB_ESPN_SCOREBOARD_JSON,
+)
+
+
+class TestMLBScraperInit:
+    """Test MLBScraper initialization."""
+
+    def test_creates_scraper_with_season(self):
+        """Test scraper initializes with correct season."""
+        scraper = MLBScraper(season=2026)
+        assert scraper.sport == "mlb"
+        assert scraper.season == 2026
+
+    def test_factory_function_creates_scraper(self):
+        """Test factory function creates correct scraper."""
+        scraper = create_mlb_scraper(season=2026)
+        assert isinstance(scraper, MLBScraper)
+        assert scraper.season == 2026
+
+    def test_expected_game_count(self):
+        """Test expected game count is correct for MLB."""
+        scraper = MLBScraper(season=2026)
+        assert scraper.expected_game_count == 2430
+
+    def test_sources_in_priority_order(self):
+        """Test sources are returned in correct priority order."""
+        scraper = MLBScraper(season=2026)
+        sources = scraper._get_sources()
+        assert sources == ["baseball_reference", "mlb_api", "espn"]
+
+
+class TestESPNParsing:
+    """Test ESPN API response parsing."""
+
+    def test_parses_completed_games(self):
+        """Test parsing completed games from ESPN."""
+        scraper = MLBScraper(season=2026)
+        data = load_json_fixture(MLB_ESPN_SCOREBOARD_JSON)
+        games = scraper._parse_espn_response(data, "http://espn.com/api")
+
+        completed = [g for g in games if g.status == "final"]
+        assert len(completed) == 2
+
+        # Yankees @ Red Sox
+        nyy_bos = next(g for g in completed if g.away_team_raw == "New York Yankees")
+        assert nyy_bos.home_team_raw == "Boston Red Sox"
+        assert nyy_bos.away_score == 3
+        assert nyy_bos.home_score == 5
+        assert nyy_bos.stadium_raw == "Fenway Park"
+
+    def test_parses_scheduled_games(self):
+        """Test parsing scheduled games from ESPN."""
+        scraper = MLBScraper(season=2026)
+        data = load_json_fixture(MLB_ESPN_SCOREBOARD_JSON)
+        games = scraper._parse_espn_response(data, "http://espn.com/api")
+
+        scheduled = [g for g in games if g.status == "scheduled"]
+        assert len(scheduled) == 1
+
+        lad_sf = scheduled[0]
+        assert lad_sf.away_team_raw == "Los Angeles Dodgers"
+        assert lad_sf.home_team_raw == "San Francisco Giants"
+        assert lad_sf.stadium_raw == "Oracle Park"
+
+    def test_parses_venue_info(self):
+        """Test venue information is extracted."""
+        scraper = MLBScraper(season=2026)
+        data = load_json_fixture(MLB_ESPN_SCOREBOARD_JSON)
+        games = scraper._parse_espn_response(data, "http://espn.com/api")
+
+        for game in games:
+            assert game.stadium_raw is not None
+
+
+class TestGameNormalization:
+    """Test game normalization and canonical ID generation."""
+
+    def test_normalizes_games_with_canonical_ids(self):
+        """Test games are normalized with correct canonical IDs."""
+        scraper = MLBScraper(season=2026)
+
+        raw_games = [
+            RawGameData(
+                game_date=datetime(2026, 4, 15),
+                home_team_raw="Boston Red Sox",
+                away_team_raw="New York Yankees",
+                stadium_raw="Fenway Park",
+                home_score=5,
+                away_score=3,
+                status="final",
+                source_url="http://example.com",
+            )
+        ]
+
+        games, review_items = scraper._normalize_games(raw_games)
+
+        assert len(games) == 1
+        game = games[0]
+
+        # Check canonical ID format
+        assert game.id == "mlb_2026_nyy_bos_0415"
+        assert game.sport == "mlb"
+        assert game.season == 2026
+
+        # Check team IDs
+        assert game.home_team_id == "team_mlb_bos"
+        assert game.away_team_id == "team_mlb_nyy"
+
+        # Check scores preserved
+        assert game.home_score == 5
+        assert game.away_score == 3
+
+    def test_creates_review_items_for_unresolved_teams(self):
+        """Test review items are created for unresolved teams."""
+        scraper = MLBScraper(season=2026)
+
+        raw_games = [
+            RawGameData(
+                game_date=datetime(2026, 4, 15),
+                home_team_raw="Unknown Team XYZ",
+                away_team_raw="Boston Red Sox",
+                stadium_raw="Fenway Park",
+                status="scheduled",
+            ),
+        ]
+
+        games, review_items = scraper._normalize_games(raw_games)
+
+        # Game should not be created due to unresolved team
+        assert len(games) == 0
+
+        # But there should be a review item
+        assert len(review_items) >= 1
+
+
+class TestTeamAndStadiumScraping:
+    """Test team and stadium data scraping."""
+
+    def test_scrapes_all_mlb_teams(self):
+        """Test all 30 MLB teams are returned."""
+        scraper = MLBScraper(season=2026)
+        teams = scraper.scrape_teams()
+
+        # 30 MLB teams
+        assert len(teams) == 30
+
+        # Check team IDs are unique
+        team_ids = [t.id for t in teams]
+        assert len(set(team_ids)) == 30
+
+        # Check all teams have required fields
+        for team in teams:
+            assert team.id.startswith("team_mlb_")
+            assert team.sport == "mlb"
+            assert team.city
+            assert team.name
+            assert team.full_name
+            assert team.abbreviation
+
+    def test_teams_have_leagues_and_divisions(self):
+        """Test teams have league (conference) and division info."""
+        scraper = MLBScraper(season=2026)
+        teams = scraper.scrape_teams()
+
+        # Count teams by league
+        al = [t for t in teams if t.conference == "American"]
+        nl = [t for t in teams if t.conference == "National"]
+
+        assert len(al) == 15
+        assert len(nl) == 15
+
+    def test_scrapes_all_mlb_stadiums(self):
+        """Test all MLB stadiums are returned."""
+        scraper = MLBScraper(season=2026)
+        stadiums = scraper.scrape_stadiums()
+
+        # Should have stadiums for all teams
+        assert len(stadiums) == 30
+
+        # Check stadium IDs are unique
+        stadium_ids = [s.id for s in stadiums]
+        assert len(set(stadium_ids)) == 30
+
+        # Check all stadiums have required fields
+        for stadium in stadiums:
+            assert stadium.id.startswith("stadium_mlb_")
+            assert stadium.sport == "mlb"
+            assert stadium.name
+            assert stadium.city
+            assert stadium.state
+            assert stadium.country in ["USA", "Canada"]
+            assert stadium.latitude != 0
+            assert stadium.longitude != 0
+
+
+class TestScrapeFallback:
+    """Test multi-source fallback behavior."""
+
+    def test_falls_back_to_next_source_on_failure(self):
+        """Test scraper tries next source when first fails."""
+        scraper = MLBScraper(season=2026)
+
+        with patch.object(scraper, '_scrape_baseball_reference') as mock_br, \
+             patch.object(scraper, '_scrape_mlb_api') as mock_mlb, \
+             patch.object(scraper, '_scrape_espn') as mock_espn:
+
+            # Make BR and MLB API fail
+            mock_br.side_effect = Exception("Connection failed")
+            mock_mlb.side_effect = Exception("API error")
+
+            # Make ESPN return data
+            mock_espn.return_value = [
+                RawGameData(
+                    game_date=datetime(2026, 4, 15),
+                    home_team_raw="Boston Red Sox",
+                    away_team_raw="New York Yankees",
+                    stadium_raw="Fenway Park",
+                    status="scheduled",
+                )
+            ]
+
+            result = scraper.scrape_games()
+
+            assert result.success
+            assert result.source == "espn"
+            assert mock_br.called
+            assert mock_mlb.called
+            assert mock_espn.called
+
+
+class TestSeasonMonths:
+    """Test season month calculation."""
+
+    def test_gets_correct_season_months(self):
+        """Test correct months are returned for MLB season."""
+        scraper = MLBScraper(season=2026)
+        months = scraper._get_season_months()
+
+        # MLB season is March-November
+        assert len(months) == 9  # Mar, Apr, May, Jun, Jul, Aug, Sep, Oct, Nov
+
+        # Check first month is March of season year
+        assert months[0] == (2026, 3)
+
+        # Check last month is November
+        assert months[-1] == (2026, 11)