hooks: run() return value controls processing of entry data.

Instead of using in-band signaling with the entry_data dict.
I don't know why I didn't think of this in the first place.

CODE.rst

@@ -51,7 +51,13 @@ Hooks make arbitrary transformations to entry data dicts.  Every entry data dict
   Initializes the hook with the user's configuration.
 
 ``run(entry_data)``
-  This method makes the hook's transformations to the entry data dict, if any.  If this method sets ``entry_data['_hook_cancel']`` to a truthy value, that entry will not be output.
+  This method can make arbitrary transformations to the entry data, or filter it so it isn't output.
+
+  If this method returns ``None``, processing the entry data continues normally.  Most hooks should do this, and just transform entry data in place.
+
+  If this method returns ``False``, processing the entry data stops immediately.  The entry will not appear in the program output.
+
+  If this method returns any other value, the program replaces the entry data with the return value, and continues processing.
 
 Templates
 ~~~~~~~~~
@@ -86,9 +92,14 @@ At a high level, import2ledger handles each input file this way::
     input_file.seek(0)
     for entry_data in importer_class(input_file):
       for hook in all_hooks:
-        hook.run(entry_data)
+        hook_return = hook.run(entry_data)
-      if entry_data:
+        if hook_return is False:
-        template.render(entry_data)
+          break
+        elif hook_return is not None:
+          entry_data = hook_return
+      else:
+        if entry_data:
+          template.render(entry_data)
 
 Note in particular that multiple importers can handle the same input file.  This helps support inputs like Patreon's earnings CSV, where completely different transactions are generated from the same source.
 

import2ledger/__main__.py

@@ -49,13 +49,15 @@ class FileImporter:
                 default_date = self.config.get_default_date()
                 in_file.seek(0)
                 for entry_data in importer(in_file):
-                    entry_data['_hook_cancel'] = False
                     for hook in self.hooks:
-                        hook.run(entry_data)
+                        hook_retval = hook.run(entry_data)
-                        if entry_data['_hook_cancel']:
+                        if hook_retval is None:
+                            pass
+                        elif hook_retval is False:
                             break
+                        else:
+                            entry_data = hook_retval
                     else:
-                        del entry_data['_hook_cancel']
                         render_vars = collections.ChainMap(entry_data, source_vars)
                         print(template.render(render_vars), file=out_file, end='')
 

import2ledger/hooks/filter_by_date.py

@@ -9,4 +9,4 @@ class FilterByDateHook:
             pass
         else:
             if not self.config.date_in_want_range(date):
-                entry_data['_hook_cancel'] = True
+                return False

tests/test_hooks.py

@@ -57,8 +57,7 @@ class DateRangeConfig:
 def test_filter_by_date(entry_date, start_date, end_date, allowed):
     entry_data = {'date': entry_date}
     hook = filter_by_date.FilterByDateHook(DateRangeConfig(start_date, end_date))
-    hook.run(entry_data)
+    assert hook.run(entry_data) is (None if allowed else False)
-    assert entry_data.get('_hook_cancel', False) == (not allowed)
 
 class DefaultDateConfig:
     ONE_DAY = datetime.timedelta(days=1)