"""
Bug checker — v2, using real static analysis instead of a heuristic.
 
Why the old approach was wrong:
The previous version fed code into a base (non-fine-tuned) CodeT5 model and
flagged "bug" whenever its generated output differed from the input. That's
not bug detection -- an untrained generative model will almost always
produce *some* different text, so it was flagging correct code constantly.
 
What this version does instead:
1. Syntax check via Python's own `ast.parse` -- catches things like missing
   colons, unmatched brackets, bad indentation. 100% accurate, since it's
   Python's real parser, not a guess.
2. Static analysis via `pyflakes` -- catches real bugs: undefined names,
   unused imports/variables, redefined functions, unreachable code, etc.
   This is the same engine that powers linters in VS Code / PyCharm.
3. (Optional) The fine-tuned seq2seq model from train.py, if you've trained
   one, is used ONLY to suggest a fix once a real issue has already been
   found by steps 1-2 -- not to decide whether a bug exists in the first
   place.
 
Install: pip install pyflakes
"""
 
import ast
import io
from pyflakes.api import check as pyflakes_check
from pyflakes.reporter import Reporter
import pycodestyle
 
 
class _ListReport(pycodestyle.BaseReport):
    """Collects pycodestyle warnings into a list instead of printing them."""
 
    def __init__(self, options):
        super().__init__(options)
        self.results = []
 
    def error(self, line_number, offset, text, check):
        code = super().error(line_number, offset, text, check)
        if code:
            self.results.append(f"line {line_number}, col {offset + 1}: {text}")
        return code
 
 
def check_style(code: str):
    """
    Runs pycodestyle (PEP8) checks.
    Returns a list of style warning strings, e.g.:
        ["line 2, col 80: E501 line too long (85 > 79 characters)"]
    Empty list means no style issues found.
    This is SEPARATE from real bugs -- style issues don't break the code,
    they just deviate from PEP8 conventions (naming, spacing, line length).
    """
    lines = code.splitlines(keepends=True)
    guide = pycodestyle.StyleGuide(reporter=_ListReport)
    report = guide.init_report()
    checker = pycodestyle.Checker(lines=lines, options=guide.options, report=report)
    try:
        checker.check_all()
    except Exception:
        return ["Style check could not process this code."]
    return report.results
 
 
 
def check_syntax(code: str):
    """Returns (is_valid: bool, error_message: str or None)"""
    try:
        ast.parse(code)
        return True, None
    except SyntaxError as e:
        return False, f"Syntax error at line {e.lineno}: {e.msg}"
 
 
def check_static_issues(code: str):
    """
    Runs pyflakes static analysis.
    Returns a list of issue strings, e.g.:
        ["line 3: undefined name 'x'", "line 1: 'os' imported but unused"]
    Empty list means no issues found.
    """
    out_stream = io.StringIO()
    err_stream = io.StringIO()
    reporter = Reporter(out_stream, err_stream)
 
    try:
        pyflakes_check(code, filename="<input>", reporter=reporter)
    except Exception:
        # pyflakes itself can't handle the code (e.g. severe syntax issues) --
        # the syntax check above should catch these first, so this is a safety net.
        return ["Static analysis could not process this code."]
 
    issues = []
    for stream in (out_stream, err_stream):
        text = stream.getvalue().strip()
        if text:
            issues.extend(line.replace("<input>:", "line ") for line in text.split("\n"))
    return issues
 
 
def detect_bugs(code: str):
    """
    Main entry point. Returns:
        {
            "has_bug": bool,
            "syntax_valid": bool,
            "syntax_error": str or None,
            "static_issues": list[str],
            "summary": str   -- human-readable summary
        }
    """
    syntax_valid, syntax_error = check_syntax(code)
 
    if not syntax_valid:
        return {
            "has_bug": True,
            "syntax_valid": False,
            "syntax_error": syntax_error,
            "static_issues": [],
            "summary": f"Syntax error found: {syntax_error}",
        }
 
    static_issues = check_static_issues(code)
    has_bug = len(static_issues) > 0
 
    if has_bug:
        summary = f"{len(static_issues)} issue(s) found via static analysis."
    else:
        summary = "No issues found by static analysis (syntax is valid, no undefined names, no unused imports/variables detected)."
 
    return {
        "has_bug": has_bug,
        "syntax_valid": True,
        "syntax_error": None,
        "static_issues": static_issues,
        "summary": summary,
    }
 
 
if __name__ == "__main__":
    # Test cases
    examples = [
        "def add(a, b)\n    return a + b",              # syntax error (missing colon)
        "def add(a, b):\n    return a + c",              # undefined name 'c'
        "import os\ndef add(a, b):\n    return a + b",   # unused import
        "def add(a, b):\n    return a + b",              # actually fine
    ]
    for code in examples:
        print("CODE:\n", code)
        print("RESULT:", detect_bugs(code))
        print("-" * 60)