Scripts/pipeline_parsers/airflow_parser.py

"""
airflow_parser.py
~~~~~~~~~~~~~~~~~
Extracts structural and quality features from Airflow DAG files.

Strategy: AST-walk the file.  We collect every Call node whose function
resolves to a known DAG constructor or task operator, plus every binary
operation (>>, <<) and explicit set_downstream / set_upstream call.

We intentionally avoid executing the file: dynamic DAG generation patterns
(DagFactory, loop-based DAGs) are flagged via heuristics rather than
skipped entirely.
"""

import ast
import json
import re
from typing import Any

# Operators we track individually because they carry specific semantic meaning
# for SE research (coupling, reliability, observability).
_SENSOR_OPERATORS = {
    "ExternalTaskSensor", "ExternalTaskMarker",
    "HttpSensor", "S3KeySensor", "SqlSensor",
    "BaseSensorOperator",
}
_BRANCH_OPERATORS = {
    "BranchPythonOperator", "BranchSQLOperator",
    "BranchDayOfWeekOperator", "BranchDateTimeOperator",
    "BaseBranchOperator",
}
_PYTHON_OPERATORS = {"PythonOperator", "PythonVirtualenvOperator", "ExternalPythonOperator"}
_BASH_OPERATORS   = {"BashOperator"}

_SCHEDULE_PRESETS = {
    "@once", "@hourly", "@daily", "@weekly", "@monthly", "@yearly", "None",
}

_CRON_RE = re.compile(
    r"^(\*|[0-9,\-\*/]+)\s+"   # minute
    r"(\*|[0-9,\-\*/]+)\s+"   # hour
    r"(\*|[0-9,\-\*/]+)\s+"   # day-of-month
    r"(\*|[0-9,\-\*/]+)\s+"   # month
    r"(\*|[0-9,\-\*/]+)$"     # day-of-week
)


def _schedule_type(raw: str | None) -> str:
    if raw is None or raw in ("None", ""):
        return "none"
    if raw in _SCHEDULE_PRESETS:
        return "preset"
    if _CRON_RE.match(raw.strip()):
        return "cron"
    if "timedelta" in raw or "datetime.timedelta" in raw:
        return "timedelta"
    if "schedule" in raw.lower() or "timetable" in raw.lower():
        return "dynamic"
    return "other"


class _DAGVisitor(ast.NodeVisitor):
    """Walks a DAG file AST and accumulates all features we care about."""

    def __init__(self) -> None:
        self.dags:               list[dict] = []
        self._current_dag:       dict | None = None
        self._operator_names:    set[str] = set()
        self._edge_count:        int = 0
        self._uses_xcom:         bool = False
        self._uses_pools:        bool = False
        self._uses_connections:  bool = False

    # ── DAG detection ─────────────────────────────────────────────────────

    def visit_With(self, node: ast.With) -> None:
        """Handles:  with DAG(...) as dag:"""
        for item in node.items:
            if isinstance(item.context_expr, ast.Call):
                name = self._call_name(item.context_expr)
                if name in ("DAG", "airflow.DAG"):
                    dag = self._parse_dag_call(item.context_expr)
                    self._current_dag = dag
                    self.generic_visit(node)
                    self._finalise_dag()
                    return
        self.generic_visit(node)

    def visit_Assign(self, node: ast.Assign) -> None:
        """Handles:  dag = DAG(...)"""
        if isinstance(node.value, ast.Call):
            name = self._call_name(node.value)
            if name in ("DAG", "airflow.DAG"):
                dag = self._parse_dag_call(node.value)
                self._current_dag = dag
                self.generic_visit(node)
                self._finalise_dag()
                return
        self.generic_visit(node)

    # ── @dag decorator ────────────────────────────────────────────────────

    def visit_FunctionDef(self, node: ast.FunctionDef) -> None:
        for dec in node.decorator_list:
            call = dec if isinstance(dec, ast.Call) else None
            bare = dec if isinstance(dec, ast.Name) else None
            name = (self._call_name(call) if call else
                    (bare.id if bare else None))
            if name == "dag":
                dag = self._parse_dag_call(call) if call else {}
                dag.setdefault("dag_id", node.name)
                self._current_dag = dag
                self.generic_visit(node)
                self._finalise_dag()
                return
        self.generic_visit(node)

    visit_AsyncFunctionDef = visit_FunctionDef

    def _finalise_dag(self) -> None:
        if self._current_dag is None:
            return
        d = self._current_dag
        d["task_count"]         = d.get("task_count", 0)
        d["edge_count"]         = self._edge_count
        d["operator_types"]     = json.dumps(sorted(self._operator_names))
        d["unique_operator_count"] = len(self._operator_names)
        d["has_python_operator"]    = int(bool(self._operator_names & _PYTHON_OPERATORS))
        d["has_bash_operator"]      = int(bool(self._operator_names & _BASH_OPERATORS))
        d["has_external_task_sensor"] = int(bool(self._operator_names & _SENSOR_OPERATORS))
        d["has_branch_operator"]    = int(bool(self._operator_names & _BRANCH_OPERATORS))
        d["uses_xcom"]           = int(self._uses_xcom)
        d["uses_pools"]          = int(self._uses_pools)
        d["uses_connections"]    = int(self._uses_connections)
        # Derive anti-patterns
        catchup        = d.get("catchup")
        max_active     = d.get("max_active_runs")
        retries        = d.get("default_retries")
        on_fail        = d.get("has_on_failure_callback", 0)
        sched_type     = d.get("schedule_type", "none")
        d["antipattern_catchup_no_maxruns"] = int(
            catchup is True and max_active is None
        )
        d["antipattern_no_retries"]   = int(retries is None or retries == 0)
        d["antipattern_no_failure_cb"] = int(not on_fail)
        d["antipattern_no_schedule"]  = int(sched_type == "none")
        d["antipattern_bare_xcom"]    = int(self._uses_xcom)  # heuristic; refine if needed
        self.dags.append(d)
        self._current_dag       = None
        self._operator_names    = set()
        self._edge_count        = 0
        self._uses_xcom         = False
        self._uses_pools        = False
        self._uses_connections  = False

    # ── Task operator detection ───────────────────────────────────────────

    def visit_Call(self, node: ast.Call) -> None:
        name = self._call_name(node)
        # Count task operators
        if name and ("Operator" in name or "Sensor" in name or "Hook" in name):
            short = name.split(".")[-1]
            self._operator_names.add(short)
            if self._current_dag is not None:
                self._current_dag["task_count"] = (
                    self._current_dag.get("task_count", 0) + 1
                )
        # xcom usage
        if name in ("xcom_push", "xcom_pull") or (
            isinstance(node.func, ast.Attribute)
            and node.func.attr in ("xcom_push", "xcom_pull")
        ):
            self._uses_xcom = True
        # pool usage
        for kw in node.keywords:
            if kw.arg == "pool":
                self._uses_pools = True
            if kw.arg in ("conn_id", "gcp_conn_id", "aws_conn_id",
                          "azure_conn_id", "http_conn_id"):
                self._uses_connections = True
        self.generic_visit(node)

    # ── Dependency edges ──────────────────────────────────────────────────

    def visit_BinOp(self, node: ast.BinOp) -> None:
        if isinstance(node.op, (ast.RShift, ast.LShift)):
            self._edge_count += 1
        self.generic_visit(node)

    def visit_Expr(self, node: ast.Expr) -> None:
        if isinstance(node.value, ast.Call):
            name = self._call_name(node.value)
            if name in ("set_downstream", "set_upstream") or (
                isinstance(node.value.func, ast.Attribute)
                and node.value.func.attr in ("set_downstream", "set_upstream")
            ):
                self._edge_count += 1
        self.generic_visit(node)

    # ── DAG constructor argument parsing ─────────────────────────────────

    def _parse_dag_call(self, node: ast.Call) -> dict:
        d: dict[str, Any] = {}
        # Positional: DAG(dag_id, schedule_interval=...)
        if node.args:
            d["dag_id"] = self._const_str(node.args[0])
        for kw in node.keywords:
            arg = kw.arg
            val = kw.value
            if arg == "dag_id":
                d["dag_id"] = self._const_str(val)
            elif arg in ("schedule_interval", "schedule"):
                raw = self._const_str(val)
                d["schedule_interval"] = raw
                d["schedule_type"] = _schedule_type(raw)
            elif arg == "catchup":
                d["catchup"] = self._const_bool(val)
            elif arg == "max_active_runs":
                d["max_active_runs"] = self._const_int(val)
            elif arg == "default_args":
                self._parse_default_args(val, d)
            elif arg == "tags":
                d["dag_tags"] = self._const_list_str(val)
            elif arg == "sla_miss_callback":
                d["has_sla"] = 1
            elif arg == "on_failure_callback":
                d["has_on_failure_callback"] = 1
            elif arg == "on_success_callback":
                d["has_on_success_callback"] = 1
        return d

    def _parse_default_args(self, node: ast.expr, d: dict) -> None:
        """Extract retries / sla / callbacks from default_args dict literal."""
        if not isinstance(node, ast.Dict):
            return
        for key, val in zip(node.keys, node.values):
            k = self._const_str(key)
            if k == "retries":
                d["default_retries"] = self._const_int(val)
            elif k == "retry_delay":
                # timedelta(minutes=...) etc. — just flag presence
                d["default_retry_delay_seconds"] = -1  # present but not trivially parsed
            elif k == "sla":
                d["has_sla"] = 1
            elif k == "on_failure_callback":
                d["has_on_failure_callback"] = 1
            elif k == "on_success_callback":
                d["has_on_success_callback"] = 1

    # ── AST helper utilities ──────────────────────────────────────────────

    @staticmethod
    def _call_name(node: ast.Call | None) -> str | None:
        if node is None:
            return None
        f = node.func
        if isinstance(f, ast.Name):
            return f.id
        if isinstance(f, ast.Attribute):
            parts = []
            cur: ast.expr = f
            while isinstance(cur, ast.Attribute):
                parts.append(cur.attr)
                cur = cur.value
            if isinstance(cur, ast.Name):
                parts.append(cur.id)
            return ".".join(reversed(parts))
        return None

    @staticmethod
    def _const_str(node: ast.expr | None) -> str | None:
        if node is None:
            return None
        if isinstance(node, ast.Constant) and isinstance(node.value, str):
            return node.value
        if isinstance(node, ast.Constant) and node.value is None:
            return "None"
        if isinstance(node, ast.Name) and node.id == "None":
            return "None"
        return repr(node)   # non-literal: store textual representation

    @staticmethod
    def _const_int(node: ast.expr | None) -> int | None:
        if isinstance(node, ast.Constant) and isinstance(node.value, int):
            return node.value
        return None

    @staticmethod
    def _const_bool(node: ast.expr | None) -> bool | None:
        if isinstance(node, ast.Constant) and isinstance(node.value, bool):
            return node.value
        if isinstance(node, ast.Name):
            if node.id == "True":
                return True
            if node.id == "False":
                return False
        return None

    @staticmethod
    def _const_list_str(node: ast.expr | None) -> str:
        if not isinstance(node, (ast.List, ast.Tuple)):
            return "[]"
        items = []
        for elt in node.elts:
            if isinstance(elt, ast.Constant) and isinstance(elt.value, str):
                items.append(elt.value)
        return json.dumps(items)


def extract(content: str, file_path: str) -> list[dict]:
    """
    Parse one Airflow DAG file and return a list of feature dicts,
    one per DAG definition found in the file.

    Returns [{"parse_error": ..., "file_lines": ...}] on parse failure.
    """
    lines = content.splitlines()
    file_lines = len(lines)
    try:
        tree = ast.parse(content, filename=file_path)
    except SyntaxError as exc:
        return [{"parse_error": str(exc), "file_lines": file_lines}]

    visitor = _DAGVisitor()
    try:
        visitor.visit(tree)
    except Exception as exc:  # noqa: BLE001
        return [{"parse_error": f"visitor: {exc}", "file_lines": file_lines}]

    if not visitor.dags:
        # No DAG found — could be a helper module in the dags/ folder.
        return [{"parse_error": "no_dag_found", "file_lines": file_lines}]

    for d in visitor.dags:
        d["file_lines"] = file_lines
        d.setdefault("parse_error", None)

    return visitor.dags