[Python-checkins] r56967 - sandbox/trunk/import_in_py/NOTES

[brett.cannon]

Author: brett.cannon
Date: Mon Aug 13 00:15:34 2007
New Revision: 56967

Modified:
   sandbox/trunk/import_in_py/NOTES
Log:
Point out how get_code() is a misplaced method by pushing code handling
requirements on to the loader when it should be done elsewhere.  Propose some
ways to deal with it.


Modified: sandbox/trunk/import_in_py/NOTES
==============================================================================
--- sandbox/trunk/import_in_py/NOTES	(original)
+++ sandbox/trunk/import_in_py/NOTES	Mon Aug 13 00:15:34 2007
@@ -4,8 +4,7 @@
     exists, raise XXX.
 * source(module_name) -> str
     Return the text for a source module.
-    XXX Can this be replaced by get_source?  Only if expected to be defined on
-    a loader.
+    XXX Consider just using get_source from PEP 302 optional extensions.
 * bytecode(module_name) -> bytes
     Return the bytes for a bytecode module.
     XXX Rename to get_bytecode if get_source is viable for source().
@@ -35,6 +34,11 @@
 * get_data(path:str) -> bytes
 * is_package(fullname:str) -> bool
 * get_code(fullname:str) -> code
+    XXX Really unfortunate as it pushes unneeded code issues into the
+    importer/loader.  For simplicity, naively compile source and fall back on
+    bytecode only if no source (require a check for the bytecode cookie?).
+    Might be prudent to end up with a function in imp that takes a loader and
+    can handle the proper thing to do based on source() and bytecode().
 * get_source(fullname:str) -> str