fix(pkg-r): Use database-agnostic SQL for SQL Server compatibility

[cpsievert]

Summary

• Replace LIMIT 0 with WHERE 1=0 to get column names (standard SQL)
• Replace LIMIT 1 with dbSendQuery() + dbFetch(n=1) for type detection sampling

Fixes #112

Problem

SQL Server uses TOP N instead of LIMIT N, causing errors like:

Incorrect syntax near '1'.
<SQL> 'SELECT * FROM "planning"."OpsAppSched" LIMIT 1'

The R package's DBISource class was using LIMIT syntax in two places:

1. initialize(): SELECT * FROM table LIMIT 0 to get column names
2. get_schema_impl(): SELECT * FROM table LIMIT 1 to sample a row for type detection

Solution

Use database-agnostic approaches that work across all SQL databases:

Fix 1: WHERE 1=0 instead of LIMIT 0

WHERE 1=0 is ANSI SQL that returns zero rows while preserving column metadata. This is a standard pattern supported by all SQL databases including SQL Server, PostgreSQL, MySQL, SQLite, and DuckDB.

Fix 2: dbFetch(n=1) instead of LIMIT 1

DBI's dbSendQuery() + dbFetch(n=1) + dbClearResult() pattern delegates row limiting to the database driver, which handles dialect-specific syntax internally. This is the same approach already used by DBISource$test_query().

Why Python doesn't need changes

The Python package handles SQL Server through different mechanisms:

1. DataFrameSource: Uses in-memory DuckDB which supports LIMIT
2. SQLAlchemySource: Already has a fallback at lines 657-662:

   try:
       result = read_sql(text(limit_query), conn)
   except Exception:
       # If LIMIT syntax doesn't work, fall back to regular read and take first row
       result = read_sql(text(query), conn).head(1)

3. PolarsLazySource: Uses Polars SQLContext which supports LIMIT
4. IbisSource: Uses result.limit(1).execute() which is Ibis's API that handles SQL dialect translation automatically

Test plan

• All R tests pass (538 tests)
• All Python tests pass (251 tests)
• Manual testing with SQL Server (not available in CI)

🤖 Generated with Claude Code

[Copilot]

Pull request overview

This PR fixes SQL Server compatibility issues in the R package's DBISource class by replacing non-standard LIMIT clauses with database-agnostic SQL patterns. The changes ensure the code works across all SQL database systems including SQL Server, which uses TOP N instead of LIMIT N.

Changes:

• Replace LIMIT 0 with WHERE 1=0 for column name retrieval
• Replace LIMIT 1 with dbSendQuery() + dbFetch(n=1) pattern for type detection sampling

---

💡 Add Copilot custom instructions for smarter, more guided reviews. Learn how to get started.

[gadenbuie]

1️⃣ 🟰 0️⃣