feat: SQL between/hostvar-key alignment, class-condition parsing, gcov merge across scenario runs

cobol_testgen/coverage.py

@@ -229,17 +229,25 @@ def _mark_if(dp, cons):
 
     simple = getattr(dp, 'parsed', None)
     if simple:
-        field, op, val = simple
-        inv_op = {'=': '<>', '<>': '=', '>': '<=', '<': '>=', '>=': '<', '<=': '>'}.get(op, op)
-        inv_simple = (field, inv_op, val)
-        for c in cons:
-            if _match_constraint(c, simple):
-                if c[3]:
-                    dp.active_branches.add('T')
-                else:
+        if len(simple) == 4:
+            # class condition (field, 'IS', CLASS, base_want):
+            # path want == base_want → T branch; == not base_want → F branch
+            field, op, val, base_want = simple
+            for c in cons:
+                if _match_constraint(c, (field, op, val)):
+                    dp.active_branches.add('T' if c[3] == base_want else 'F')
+        else:
+            field, op, val = simple
+            inv_op = {'=': '<>', '<>': '=', '>': '<=', '<': '>=', '>=': '<', '<=': '>'}.get(op, op)
+            inv_simple = (field, inv_op, val)
+            for c in cons:
+                if _match_constraint(c, simple):
+                    if c[3]:
+                        dp.active_branches.add('T')
+                    else:
+                        dp.active_branches.add('F')
+                elif _match_constraint(c, inv_simple):
                     dp.active_branches.add('F')
-            elif _match_constraint(c, inv_simple):
-                dp.active_branches.add('F')
     elif dp.cond_tree and dp.cond_leaves:
         assignment = {}
         for leaf in dp.cond_leaves:
@@ -411,17 +419,24 @@ def _mark_perform(dp, cons):
 
     simple = getattr(dp, 'parsed', None)
     if simple:
-        field, op, val = simple
-        inv_op = {'=': '<>', '<>': '=', '>': '<=', '<': '>=', '>=': '<', '<=': '>'}.get(op, op)
-        inv_simple = (field, inv_op, val)
-        for c in cons:
-            if _match_constraint(c, simple):
-                if c[3]:
-                    dp.active_branches.add('Skip')
-                else:
+        if len(simple) == 4:
+            # class condition (field, 'IS', CLASS, base_want): path want == base_want → Skip
+            field, op, val, base_want = simple
+            for c in cons:
+                if _match_constraint(c, (field, op, val)):
+                    dp.active_branches.add('Skip' if c[3] == base_want else 'Enter')
+        else:
+            field, op, val = simple
+            inv_op = {'=': '<>', '<>': '=', '>': '<=', '<': '>=', '>=': '<', '<=': '>'}.get(op, op)
+            inv_simple = (field, inv_op, val)
+            for c in cons:
+                if _match_constraint(c, simple):
+                    if c[3]:
+                        dp.active_branches.add('Skip')
+                    else:
+                        dp.active_branches.add('Enter')
+                elif _match_constraint(c, inv_simple):
                     dp.active_branches.add('Enter')
-            elif _match_constraint(c, inv_simple):
-                dp.active_branches.add('Enter')
     elif dp.cond_tree and dp.cond_leaves:
         assignment = {}
         for leaf in dp.cond_leaves:
@@ -482,7 +497,42 @@ def locate_decision_lines(decision_points, raw_source):
                     if re.search(short_pat, lines[i]):
                         dp.source_line = i + 1
                         used_indices[dp.label] = i
+                        found = True
                         break
+        # Multi-line fallback 2: 条件被换行拆分（如 IF MERGE-REC-TYPE / = CONST）
+        # 用条件首个字段定位 IF 起始行（gcov 分支标记需要源行）
+        if not found and dp.kind == 'IF':
+            first_field_pat = _build_first_field_if_pattern(dp)
+            if first_field_pat:
+                for i in range(start, len(lines)):
+                    if re.search(first_field_pat, lines[i]):
+                        dp.source_line = i + 1
+                        used_indices[dp.label] = i
+                        break
+
+
+def _build_first_field_if_pattern(dp):
+    """Build a pattern matching 'IF <首字段>' — 用于条件被换行拆分的 IF。
+
+    例：条件 'MERGE-REC-TYPE = CNS-PAY-TYPE-SALARY' 在源码中写作
+        IF MERGE-REC-TYPE
+            = CNS-PAY-TYPE-SALARY
+    单行模式无法匹配；此模式仅匹配 'IF MERGE-REC-TYPE' 定位起始行。
+    """
+    if dp.kind != 'IF':
+        return None
+    label = dp.label or ''
+    cond = label[2:].strip() if label.upper().startswith('IF ') else label
+    cond = cond.strip()
+    if not cond:
+        return None
+    # 去除前导 NOT / 括号，取首字段名
+    m = re.match(r"^(?:NOT\s+)?\(?([A-Z][A-Z0-9_-]*)", cond, re.IGNORECASE)
+    if not m:
+        return None
+    first = m.group(1)
+    esc = re.escape(first)
+    return r'\bIF\b\s+' + esc + r'\b'
 
 
 def _normalize(text):
@@ -494,7 +544,11 @@ def _normalize(text):
 def _build_search_patterns(dp):
     texts = []
     if dp.kind == 'IF':
-        texts.append((r'\bIF\b', dp.label))
+        # dp.label is already like "IF SQLCODE = -803" — avoid emitting
+        # "\bIF\b\s+IF\s+..." which never matches a real source line.
+        label = dp.label or ''
+        cond = label[2:].strip() if label.upper().startswith('IF ') else label
+        texts.append((r'\bIF\b', cond))
     elif dp.kind == 'EVALUATE':
         texts.append((r'\bEVALUATE\b', dp.label))
     elif dp.kind == 'PERFORM':