File size: 8,651 Bytes
893a23c
 
009dca3
893a23c
 
 
 
 
009dca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
893a23c
 
009dca3
893a23c
009dca3
893a23c
009dca3
893a23c
009dca3
 
893a23c
009dca3
893a23c
009dca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
893a23c
009dca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
893a23c
 
 
009dca3
 
893a23c
009dca3
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
 
893a23c
 
 
009dca3
893a23c
 
 
009dca3
 
 
 
 
 
 
 
 
1
2
3
4
5
6
7
8
9
10
11
12
13
14
15
16
17
18
19
20
21
22
23
24
25
26
27
28
29
30
31
32
33
34
35
36
37
38
39
40
41
42
43
44
45
46
47
48
49
50
51
52
53
54
55
56
57
58
59
60
61
62
63
64
65
66
67
68
69
70
71
72
73
74
75
76
77
78
79
80
81
82
83
84
85
86
87
88
89
90
91
92
93
94
95
96
97
98
99
100
101
102
103
104
105
106
107
108
109
110
111
112
113
114
115
116
117
118
119
120
121
122
123
124
125
126
127
128
129
130
131
132
133
134
135
136
137
138
139
140
141
142
143
144
145
146
147
148
149
150
151
152
153
154
155
156
157
158
159
160
161
162
163
164
165
166
167
168
169
170
171
172
173
174
175
176
177
178
179
180
181
182
183
184
185
186
187
188
189
190
191
192
193
194
195
196
197
198
199
200
201
202
203
204
205
206
207
208
209
210
211
212
213
214
215
216
217
218
219
220
221
222
223
224
225
226
227
228
229
230
231
232
233
234
235
236
237
238
239
240
241
242
243
244
245
"""
executor.py — Pure-Python ETL pipeline interpreter.
Supports step types: cast, select, rename, filter, date_parse, join, agg, dedup
"""
import copy
import datetime
from typing import Any

# --- Execution Engine ---

def execute_step(rows: list[dict], step: dict) -> list[dict]:
    """Execute a single pipeline step against a list of rows."""
    op = step.get("op")
    field = step.get("field")
    
    if op == "cast":
        to_type = step.get("to_type")
        null_handling = step.get("null_handling", "error") # error | coerce | drop
        return _op_cast(rows, field, to_type, null_handling)
    
    elif op == "select":
        columns = step.get("columns", [])
        return _op_select(rows, columns)
    
    elif op == "rename":
        mapping = step.get("mapping", {})
        return _op_rename(rows, mapping)
    
    elif op == "filter":
        condition = step.get("condition")
        value = step.get("value")
        return _op_filter(rows, field, condition, value)
    
    elif op == "dedup":
        subset = step.get("subset", [])
        keep = step.get("keep", "first")
        return _op_dedup(rows, subset, keep)
    
    elif op == "join":
        join_type = step.get("join_type", "left")
        on = step.get("on")
        right = step.get("right", [])
        return _op_join(rows, on, right, join_type)
    
    elif op == "agg":
        group_by = step.get("group_by", [])
        aggregations = step.get("aggregations", {})
        return _op_agg(rows, group_by, aggregations)
    
    else:
        raise ValueError(f"Unknown operation: {op!r}")

def run_pipeline(config: dict, rows: list[dict]) -> tuple[list[dict], str | None]:
    """Execute a sequence of steps defined in config['steps']."""
    data = copy.deepcopy(rows)
    for i, step in enumerate(config.get("steps", [])):
        try:
            data = execute_step(data, step)
        except Exception as e:
            # Warm Baseline: Return current data (last valid state) instead of empty list
            return data, f"Step {i} ({step.get('op')}): {type(e).__name__}: {e}"

    return data, None

# --- Internal Ops ---

def _op_cast(rows, field, to_type, null_handling):
    out = []
    for row in rows:
        val = row.get(field)
        
        # Check for null-like values
        is_null = val is None or str(val).lower().strip() in ("none", "null", "n/a", "")
        
        if is_null:
            if null_handling == "coerce":
                out.append({**row, field: None})
            elif null_handling == "drop":
                continue
            else:
                raise ValueError(f"invalid input {val!r} for type {to_type}")
            continue

        try:
            if to_type == "DATE":
                # Basic YYYY-MM-DD check, but keep logic flexible for tests
                # The prompt tests expect "2024-01-15" to pass and "2024-99-99" to fail
                try:
                    # Try parsing to validate
                    datetime.datetime.strptime(str(val).strip(), "%Y-%m-%d")
                    out.append({**row, field: str(val).strip()})
                except ValueError:
                    raise ValueError(f"invalid date {val!r}")
            elif to_type == "INT":
                out.append({**row, field: int(float(str(val)))})
            elif to_type == "FLOAT":
                out.append({**row, field: float(str(val))})
            elif to_type == "STRING":
                out.append({**row, field: str(val)})
            elif to_type == "BOOLEAN":
                out.append({**row, field: str(val).lower() in ("1", "true", "yes")})
            else:
                out.append({**row, field: val})
        except Exception as e:
            if null_handling == "coerce":
                out.append({**row, field: None})
            elif null_handling == "drop":
                continue
            else:
                raise ValueError(f"invalid input {val!r} for type {to_type}: {e}")
    return out

def _op_select(rows, columns):
    return [{col: row[col] for col in columns} for row in rows]

def _op_rename(rows, mapping):
    return [{mapping.get(k, k): v for k, v in row.items()} for row in rows]

def _op_filter(rows, field, condition, value):
    def check(val):
        if condition == "eq": return val == value
        if condition == "not_null": return val is not None
        if condition == "is_null": return val is None
        return True
    return [row for row in rows if check(row.get(field))]

def _op_dedup(rows, subset, keep):
    seen = set()
    out = []
    for row in rows:
        key = tuple(row.get(s) for s in subset)
        if key not in seen:
            out.append(row)
            seen.add(key)
        elif keep == "all": # Dummy for completeness
            out.append(row)
    return out

def _op_join(rows, on, right, join_type):
    # Simplified left/inner join
    right_lookup = {}
    from collections import defaultdict
    multi_lookup = defaultdict(list)
    for r in right:
        k = r.get(on)
        multi_lookup[k].append(r)
        if k not in right_lookup:
            right_lookup[k] = r
            
    out = []
    for row in rows:
        k = row.get(on)
        matches = multi_lookup.get(k)
        if matches:
            for m in matches:
                merged = {**row}
                for mk, mv in m.items():
                    if mk != on: merged[mk] = mv
                out.append(merged)
        else:
            if join_type == "left":
                merged = {**row}
                # Find all potential keys from right to fill with None
                if right:
                    for rk in right[0].keys():
                        if rk != on and rk not in merged:
                            merged[rk] = None
                out.append(merged)
    return out

def _op_agg(rows, group_by, aggregations):
    from collections import defaultdict
    groups = defaultdict(list)
    for row in rows:
        key = tuple(row.get(g) for g in group_by)
        groups[key].append(row)
        
    out = []
    for key, grouped_rows in groups.items():
        base = {g: k for g, k in zip(group_by, key)}
        for key, agg_spec in aggregations.items():
            field = agg_spec["field"]
            func = agg_spec["func"]
            # Naming Logic: use explicit output_name if provided, else fallback to default
            out_name = agg_spec.get("output_name", f"{field}_{func}")
            
            vals = [r.get(field) for r in grouped_rows if r.get(field) is not None]
            
            if func == "sum":
                base[out_name] = sum(vals)
            elif func == "count_distinct":
                base[out_name] = len(set(vals))
            elif func == "count":
                base[out_name] = len(vals)
            elif func == "mean":
                base[out_name] = sum(vals)/len(vals) if vals else None

                
        out.append(base)
    return out

# --- Utility ---

def compare_output(got: list[dict], expected: list[dict]) -> dict:
    if not expected:
        return {"row_match": 1.0 if not got else 0.0, "schema_match": not got}
    
    schema_match = bool(got) and set(got[0].keys()) == set(expected[0].keys())
    
    matching_rows = 0
    # Warm Baseline: If schema fails, check if the VALUES match (flexible row matching)
    for g, e in zip(got, expected):
        if g == e:
            matching_rows += 1
        elif not schema_match:
            # Check if values match regardless of keys (for silent schema mismatches)
            try:
                g_vals = sorted(str(round(float(v), 3)) if isinstance(v, (int, float)) else str(v) for v in g.values())
                e_vals = sorted(str(round(float(v), 3)) if isinstance(v, (int, float)) else str(v) for v in e.values())
                if g_vals == e_vals:
                    matching_rows += 0.8 # 80% credit for correct values but wrong keys
            except:
                if sorted(str(v) for v in g.values()) == sorted(str(v) for v in e.values()):
                    matching_rows += 0.8

        
    row_match = matching_rows / len(expected) if expected else 1.0
    
    return {
        "row_match": round(row_match, 4),
        "schema_match": schema_match,
        "exact_match": row_match == 1.0 and schema_match
    }


def apply_patch(config: dict, patch: dict) -> dict:
    config = copy.deepcopy(config)
    step_idx = patch["step_index"]
    field = patch["field"]
    new_val = patch["new_value"]
    
    # Flat schema: config["steps"][step_idx][field] = new_val
    config["steps"][step_idx][field] = new_val
    return config