Merge pull request #2810 from stfc/martin_fileinfo_cache

Add file caching features to FileInfo

.gitignore

@@ -29,3 +29,5 @@ src/*.egg-info
 .venv
 cov.xml
 .coverage.*
+*.psycache
+__pycache__

changelog

@@ -26,6 +26,9 @@
 
 	9) PR #2842 for #2841. Update all copyright date ranges for 2025.
 
+	10) PR #2810. Adds caching of the fparser2 parse tree to FileInfo. Is
+	disabled by default.
+
 release 3.0.0 6th of December 2024
 
 	1) PR #2477 for #2463. Add support for Fortran Namelist statements.

doc/developer_guide/module_manager.rst

@@ -41,6 +41,8 @@
 
 .. _module_manager:
 
+
+
 Module Manager
 ##############
 
@@ -135,3 +137,65 @@ which prints the filenames of all modules used in ``tl_testkern_mod``:
     Module: constants_mod constants_mod.f90
     Module: fs_continuity_mod fs_continuity_mod.f90
     Module: kernel_mod kernel_mod.f90
+
+
+
+FileInfo
+========
+
+FileInfo is a class that is used to store information about Fortran files.
+
+This information can include:
+
+- The source code itself
+- The fparser tree information
+- The PSyIR tree information
+
+All this information is gathered in this single class since this also
+allows for caching of it, see next section
+
+
+
+Caching
+=======
+
+The `ModuleManager` and `FileInfo` support a caching of the
+fparser tree representation of a source code.
+(Support for PSyIR is planned)
+
+This caching has to be **explicitly enabled** in the constructor
+of `ModuleManager`.
+
+
+.. testcode ::
+
+    mod_manager = ModuleManager.get(use_caching=True)
+
+
+Most of the time in the PSyIR generation is currently spent in the
+fparser tree generation. Consequently, this leads to significant
+speed-ups in the process of reading and parsing the source code
+of modules.
+
+The default cache file is named the same way as the source file,
+but replaces the file extension with `.psycache`. E.g., a cache file
+for the source file `foo.f90` will be called `foo.psycache`.
+
+The caching algorithm to obtain the fparser tree OR PSyIR is briefly described as follows:
+
+- If fparser tree / PSyIR was read before: RETURN fparser tree or PSyIR
+- If source code is not yet read:
+
+    - Read the content of the file
+    - Create the source's checksum.
+- Read cache file if it exists:
+
+    - If the checksum of the cache is the same as the one of the source:
+
+        - load the fparser tree / PSyIR from the cache file and RETURN fparser tree or PSyIR
+- Create the fparser tree / PSyIR from the source code
+- Save cache file IF it was not loaded before:
+
+    - Update cache information
+    - Store to cache file
+- RETURN fparser tree or PSyIR

doc/user_guide/configuration.rst

@@ -76,6 +76,10 @@ locations searched is now:
 2. ``<python-base-dir>/share/psyclone/``
 3. ``${HOME}/.local/share/psyclone/``
 
+As a last resort, the location
+``<psyclone-src-base>/config/``
+is searched in case PSyclone was installed in editable mode.
+
 Note that for developers a slightly different configuration handling
 is implemented, see :ref:`dev_guide:dev_configuration` for details.
 

src/psyclone/configuration.py

@@ -415,6 +415,7 @@ def find_file():
         - ${HOME}/.local/share/psyclone/
         - <system-install-prefix>/share/psyclone/
         - <psyclone-installation-base>/share/psyclone/
+        - <psyclone-src-base>/config/
 
         :returns: the fully-qualified path to the configuration file
         :rtype: str
@@ -449,9 +450,18 @@ def find_file():
         if not within_virtual_env():
             # 4. <python-installation-base>/share/psyclone/
             _file_paths.append(share_dir)
+
         # 5. <psyclone-installation-base>/share/psyclone/
         _file_paths.extend(pkg_share_dir)
 
+        # 6. <psyclone-src-base>/config/
+        # Search for configuration file relative to this source file
+        dev_dir_tmp = os.path.dirname(__file__)
+        dev_dir_tmp = os.path.split(dev_dir_tmp)[0]  # Go down one level
+        dev_dir_tmp = os.path.split(dev_dir_tmp)[0]  # Go down another level
+        dev_path = os.path.join(dev_dir_tmp, "config")
+        _file_paths.append(dev_path)
+
         for cfile in [os.path.join(cdir, _FILE_NAME) for cdir in _file_paths]:
             if os.path.isfile(cfile):
                 return cfile

src/psyclone/parse/__init__.py

@@ -32,18 +32,20 @@
 # POSSIBILITY OF SUCH DAMAGE.
 # -----------------------------------------------------------------------------
 # Author: J. Henrichs, Bureau of Meteorology
+# Modified: M. Schreiber, Univ. Grenoble Alpes / Inria / Lab. Jean-Kuntzmann
 
 '''This directory contains classes related to parsing Fortran.
 '''
 
-from psyclone.parse.file_info import FileInfo
+from psyclone.parse.file_info import FileInfo, FileInfoFParserError
 from psyclone.parse.module_info import ModuleInfo, ModuleInfoError
 from psyclone.parse.module_manager import ModuleManager
 
 
 # For AutoAPI documentation generation.
 __all__ = [
         'FileInfo',
+        'FileInfoFParserError',
         'ModuleInfo',
         'ModuleInfoError',
         'ModuleManager'