Merge pull request #187 from ecmwf-ifs/186-empty-file-regex-frontend

Fix handling of empty files in frontends (fix #186)

loki/backend/fgen.py

@@ -904,7 +904,7 @@ def fgen(ir, depth=0, conservative=False, linewidth=132):
     """
     Generate standardized Fortran code from one or many IR objects/trees.
     """
-    return FortranCodegen(depth=depth, linewidth=linewidth, conservative=conservative).visit(ir)
+    return FortranCodegen(depth=depth, linewidth=linewidth, conservative=conservative).visit(ir) or ''
 
 
 """

loki/frontend/fparser.py

@@ -268,7 +268,7 @@ def get_source(self, o, source):
         """
         Helper method that builds the source object for the node.
         """
-        if not isinstance(o, str) and o.item is not None:
+        if o is not None and not isinstance(o, str) and o.item is not None:
             lines = (o.item.span[0], o.item.span[1])
             string = ''.join(self.raw_source[lines[0] - 1:lines[1]]).strip('\n')
             source = Source(lines=lines, string=string)
@@ -288,7 +288,7 @@ def get_label(self, o):
         """
         Helper method that returns the label of the node.
         """
-        if not isinstance(o, str) and o.item is not None:
+        if o is not None and not isinstance(o, str) and o.item is not None:
             return getattr(o.item, 'label', None)
         return None
 

loki/frontend/source.py

@@ -243,7 +243,10 @@ def to_source(self, include_padding=False):
         """
         Create a :any:`Source` object with the content of the reader
         """
-        if include_padding:
+        if not self.source_lines:
+            string = ''
+            lines = (self.line_offset + 1, self.line_offset + 1)
+        elif include_padding:
             string = '\n'.join(self.source_lines)
             lines = (self.line_offset + 1, self.line_offset + len(self.source_lines))
         else:

tests/test_frontends.py

@@ -13,7 +13,7 @@
     Deallocation, Associate, BasicType, OMNI, OFP, FP, Enumeration,
     config, REGEX, Sourcefile, Import, RawSource, CallStatement,
     RegexParserClass, ProcedureType, DerivedType, Comment, Pragma,
-    PreprocessorDirective, config_override
+    PreprocessorDirective, config_override, Section
 )
 from loki.expression import symbols as sym
 
@@ -134,9 +134,7 @@ def test_check_alloc_opts(here, frontend):
 
 @pytest.mark.parametrize('frontend', available_frontends())
 def test_associates(here, frontend):
-    """
-    Test the use of associate to access and modify other items
-    """
+    """Test the use of associate to access and modify other items"""
 
     fcode = """
 module derived_types_mod
@@ -255,9 +253,7 @@ def test_associates_deferred(frontend):
 
 @pytest.mark.parametrize('frontend', available_frontends())
 def test_associates_expr(here, frontend):
-    """
-    Verify that associates with expressions are supported
-    """
+    """Verify that associates with expressions are supported"""
     fcode = """
 subroutine associates_expr(in, out)
   implicit none
@@ -296,9 +292,7 @@ def test_associates_expr(here, frontend):
 
 @pytest.mark.parametrize('frontend', available_frontends())
 def test_enum(here, frontend):
-    """
-    Verify that enums are represented correctly
-    """
+    """Verify that enums are represented correctly"""
     # F2008, Note 4.67
     fcode = """
 subroutine test_enum (out)
@@ -1490,3 +1484,15 @@ def test_regex_fypp():
     assert len(module.routines) == 2
     assert module.routines[0].name == 'first_routine'
     assert module.routines[1].name == 'last_routine'
+
+
+@pytest.mark.parametrize(
+    'frontend', 
+    available_frontends(include_regex=True, xfail=[(OMNI, 'OMNI may segfault on empty files')])
+)
+@pytest.mark.parametrize('fcode', ['', '\n', '\n\n\n\n'])
+def test_frontend_empty_file(frontend, fcode):
+    """Ensure that all frontends can handle empty source files correctly (#186)"""
+    source = Sourcefile.from_source(fcode, frontend=frontend)
+    assert isinstance(source.ir, Section)
+    assert not source.to_fortran().strip()

tests/test_source.py

@@ -452,3 +452,16 @@ def sanitize_empty_lines_and_comments(string):
     for _ in reader:
         iterated_code += reader.source_from_current_line().string + '\n'
     assert sanitize_empty_lines_and_comments(fcode) == iterated_code
+
+
+@pytest.mark.parametrize('fcode', ['', '\n'])
+def test_fortran_reader_empty(fcode):
+    """Test :any:`FortranReader` for empty strings"""
+    reader = FortranReader(fcode)
+    assert isinstance(reader, FortranReader)
+    assert not reader.source_lines
+    assert not reader.sanitized_lines
+    source = reader.to_source()
+    assert isinstance(source, Source)
+    assert source.lines == (1, 1)
+    assert source.string == ''