Add C support for local string declarations (pyccel#2064)

Add C support for local string declarations. This is a first step
towards allowing string keys for dictionaries. Fixes pyccel#2061

**Commit Summary**
- Handle strings in loop unravelling
- Allow extensions to depend on stc folders by using a recognisable
dependency that can be swapped for the correct directory
- Fix declaration of strings in C
- Add support for `PyccelArrayShapeElement` of a string in C
- Remove double line break after deallocation
- Add support for string initialisation
- Correct the warning message for array reallocation (maybe this should
be removed for non-arrays)
- Ensure strings are allocated in the semantic parser.
- Create a `cstr.c` file to compile string functions independently
(avoids duplicate definitions)
- Activate tests
- Add a leak test for string reallocation

Author: EmilyBourne

CHANGELOG.md

@@ -55,6 +55,7 @@ All notable changes to this project will be documented in this file.
 -   #2016 : Add support for translating arithmetic magic methods (methods cannot yet be used from Python).
 -   #1980 : Extend The C support for min and max to more than two variables
 -   #2081 : Add support for multi operator expressions
+-   #2061 : Add C support for string declarations.
 -   Add support for inhomogeneous tuple annotations.
 -   #1834 : Add support for `@property` decorator.
 -   \[INTERNALS\] Add abstract class `SetMethod` to handle calls to various set methods.

pyccel/ast/utilities.py

@@ -425,7 +425,12 @@ def collect_loops(block, indices, new_index, language_has_vectors = False, resul
                                     and not isinstance(f, (NumpyTranspose))))
     for line in block:
 
-        if (isinstance(line, Assign) and
+        if isinstance(line, Assign) and isinstance(line.lhs.class_type, StringType):
+            # Save line in top level (no for loop)
+            result.append(line)
+            current_level = 0
+
+        elif (isinstance(line, Assign) and
                 not isinstance(line.rhs, (array_creator_types, Nil)) and # not creating array
                 not line.rhs.get_attribute_nodes(array_creator_types) and # not creating array
                 not is_function_call(line.rhs)): # not a basic function call

pyccel/codegen/compiling/basic.py

@@ -91,10 +91,7 @@ def __init__(self,
         self._lock_source  = FileLock(self.source+'.lock')
 
         self._flags        = list(flags)
-        if has_target_file:
-            self._includes     = set([folder, *includes])
-        else:
-            self._includes = set(includes)
+        self._includes     = set([folder, *includes])
         self._libs         = list(libs)
         self._libdirs      = set(libdirs)
         self._accelerators = set(accelerators)
@@ -117,9 +114,8 @@ def reset_folder(self, folder):
         folder : str
             The new folder where the source file can be found.
         """
-        if self.has_target_file:
-            self._includes.remove(self._folder)
-            self._includes.add(folder)
+        self._includes.remove(self._folder)
+        self._includes.add(folder)
 
         self._file = os.path.join(folder, os.path.basename(self._file))
         self._lock_source  = FileLock(self.source+'.lock')

pyccel/codegen/pipeline.py

@@ -311,7 +311,7 @@ def handle_error(stage):
     #         pass
     #------------------------------------------------------
     try:
-        manage_dependencies(codegen.printer, src_compiler, pyccel_dirpath, mod_obj,
+        manage_dependencies(codegen.printer.get_additional_imports(), src_compiler, pyccel_dirpath, mod_obj,
                 language, verbose, convert_only)
     except NotImplementedError as error:
         errors.report(f'{error}\n'+PYCCEL_RESTRICTION_TODO,

pyccel/codegen/printing/ccode.py

@@ -247,7 +247,8 @@
                  'stdio',
                  "inttypes",
                  'stdbool',
-                 'assert']}
+                 'assert',
+                 'stc/cstr']}
 
 import_header_guard_prefix = {'stc/hset'    : '_TOOLS_SET',
                               'stc/vec'   : '_TOOLS_LIST',
@@ -309,7 +310,6 @@ class CCodePrinter(CodePrinter):
                       (PrimitiveIntegerType(),8)       : LiteralString("%") + CMacro('PRId64'),
                       (PrimitiveIntegerType(),2)       : LiteralString("%") + CMacro('PRId16'),
                       (PrimitiveIntegerType(),1)       : LiteralString("%") + CMacro('PRId8'),
-                      StringType()                  : '%s',
                       }
 
     def __init__(self, filename, prefix_module = None):
@@ -1039,7 +1039,7 @@ def _print_Import(self, expr):
                      f'#include <{stc_extension_mapping[source]}.h>\n', 
                      f'#endif // {header_guard}\n\n'))
             return code
-        elif source.startswith('stc/'):
+        elif source != 'stc/cstr' and (source.startswith('stc/') or source in import_header_guard_prefix):
             code = ''
             for t in expr.target:
                 class_type = t.object.class_type
@@ -1130,6 +1130,13 @@ def get_print_format_and_arg(self, var):
                 except KeyError:
                     errors.report(f"Printing {var.dtype} type is not supported currently", severity='fatal')
                 arg = self._print(var)
+        elif isinstance(var.dtype, StringType):
+            if isinstance(var, Variable):
+                var_obj = self._print(ObjectAddress(var))
+                arg = f'cstr_str({var_obj})'
+            else:
+                arg = self._print(var)
+            arg_format = '%s'
         else:
             try:
                 arg_format = self.type_to_format[var.dtype]
@@ -1302,6 +1309,9 @@ def get_c_type(self, dtype):
             i_type = f'{container_type}_{key_type}_{val_type}'
             self.add_import(Import(f'stc/{container_type}', AsName(VariableTypeAnnotation(dtype), i_type)))
             return i_type
+        elif isinstance(dtype, StringType):
+            self.add_import(c_imports['stc/cstr'])
+            return 'cstr'
         else:
             key = dtype
 
@@ -1377,11 +1387,11 @@ def get_declare_type(self, expr):
         rank  = expr.rank
 
         if rank > 0:
-            if isinstance(expr.class_type, (HomogeneousSetType, HomogeneousListType, DictType)):
-                dtype = self.get_c_type(expr.class_type)
-            elif isinstance(expr.class_type, CStackArray):
-                return self.get_c_type(expr.class_type.element_type)
-            elif isinstance(expr.class_type, (HomogeneousTupleType, NumpyNDArrayType)):
+            if isinstance(class_type, (HomogeneousSetType, HomogeneousListType, DictType, StringType)):
+                dtype = self.get_c_type(class_type)
+            elif isinstance(class_type, CStackArray):
+                return self.get_c_type(class_type.element_type)
+            elif isinstance(class_type, (HomogeneousTupleType, NumpyNDArrayType)):
                 if expr.rank > 15:
                     errors.report(UNSUPPORTED_ARRAY_RANK, symbol=expr, severity='fatal')
                 self.add_import(c_imports['ndarrays'])
@@ -1699,15 +1709,18 @@ def _print_PyccelArrayShapeElement(self, expr):
             c_type = self.get_c_type(arg.class_type)
             arg_code = self._print(ObjectAddress(arg))
             return f'{c_type}_size({arg_code})'
+        elif isinstance(arg.class_type, StringType):
+            arg_code = self._print(ObjectAddress(arg))
+            return f'cstr_size({arg_code})'
         else:
             raise NotImplementedError(f"Don't know how to represent shape of object of type {arg.class_type}")
 
     def _print_Allocate(self, expr):
         free_code = ''
         variable = expr.variable
-        if isinstance(variable.class_type, (HomogeneousListType, HomogeneousSetType, DictType)):
+        if isinstance(variable.class_type, (HomogeneousListType, HomogeneousSetType, DictType, StringType)):
             if expr.status in ('allocated', 'unknown'):
-                free_code = f'{self._print(Deallocate(variable))}\n'
+                free_code = f'{self._print(Deallocate(variable))}'
             if expr.shape[0] is None:
                 return free_code
             size = self._print(expr.shape[0])
@@ -1718,7 +1731,7 @@ def _print_Allocate(self, expr):
             elif expr.alloc_type == 'resize':
                 return f'{container_type}_resize({variable_address}, {size}, {0});\n'
             return free_code
-        if isinstance(variable.class_type, (NumpyNDArrayType, HomogeneousTupleType)):
+        elif isinstance(variable.class_type, (NumpyNDArrayType, HomogeneousTupleType)):
             #free the array if its already allocated and checking if its not null if the status is unknown
             if  (expr.status == 'unknown'):
                 shape_var = DottedVariable(VoidType(), 'shape', lhs = variable)
@@ -1755,7 +1768,7 @@ def _print_Allocate(self, expr):
             raise NotImplementedError(f"Allocate not implemented for {variable.class_type}")
 
     def _print_Deallocate(self, expr):
-        if isinstance(expr.variable.class_type, (HomogeneousListType, HomogeneousSetType, DictType)):
+        if isinstance(expr.variable.class_type, (HomogeneousListType, HomogeneousSetType, DictType, StringType)):
             if expr.variable.is_alias:
                 return ''
             variable_address = self._print(ObjectAddress(expr.variable))
@@ -2306,11 +2319,13 @@ def _print_Assign(self, expr):
             return self.copy_NumpyArray_Data(expr)
         if isinstance(rhs, (NumpyFull)):
             return self.arrayFill(expr)
-        lhs = self._print(expr.lhs)
+        lhs_code = self._print(lhs)
         if isinstance(rhs, (PythonList, PythonSet, PythonDict)):
             return self.init_stc_container(rhs, expr)
-        rhs = self._print(expr.rhs)
-        return f'{lhs} = {rhs};\n'
+        rhs_code = self._print(rhs)
+        if isinstance(rhs, LiteralString):
+            rhs_code = f'cstr_lit({rhs_code})'
+        return f'{lhs_code} = {rhs_code};\n'
 
     def _print_AliasAssign(self, expr):
         lhs_var = expr.lhs

pyccel/codegen/python_wrapper.py

@@ -198,7 +198,7 @@ def create_shared_library(codegen,
     #  Compile cwrapper_ndarrays from stdlib (if necessary)
     #--------------------------------------------------------
     start_compile_libs = time.time()
-    manage_dependencies(wrapper_codegen, wrapper_compiler, pyccel_dirpath, wrapper_compile_obj,
+    manage_dependencies(wrapper_codegen.get_additional_imports(), wrapper_compiler, pyccel_dirpath, wrapper_compile_obj,
             language, verbose)
     timings['Dependency compilation'] += (time.time() - start_compile_libs)