Change the order in which patterns are run.

Previously patterns have been run in the order they are defined in the
pattern file with general patterns run first, then script patterns,
then language patterns and finally country patterns. Change this so
that patterns with the same name are always run one after the other in
the position of the earliest one.

For example, consider the following patterns defined in the pattern
files in the following order.

    Latn.common-error:    A, B, C
    Latn-fi.common-error: B, D, E

Previously these were run in order A, B, C, B, D, E. With this change
they will now be run in order A, B, B, C, D, E or if the policy of the
B pattern in Latn-fi has been set to 'Replace', the A, B, C, D, E.
This should make it more convenient to replace or extend patterns
above in the hierarchy.

Adapt Latn-fr.common-error.in to the above order and add
Latn-fi.common-error.in to use the new order.

data/patterns/Latn-fi.common-error.conf

@@ -0,0 +1,4 @@
+<?xml version="1.0" encoding="utf-8"?>
+<patterns>
+  <pattern name="Space after punctuation marks" enabled="true"/>
+</patterns>

data/patterns/Latn-fi.common-error.in

@@ -0,0 +1,11 @@
+[Common Error Pattern]
+_Name=Space after punctuation marks
+_Description=Add space after various punctuation marks
+Classes=Human;OCR;
+# Same as the Latin pattern, except no space after colon to accommodate
+# Finnish suffixes, e.g. 'TV:ssä'.
+Pattern=((\w|^|["'«»])[,;?!])(?!["'«»])([^\W\d][\w\s])
+Flags=DOTALL;MULTILINE;UNICODE;
+Replacement=\1 \3
+Repeat=False
+Policy=Replace

data/patterns/Latn-fr.common-error.in

@@ -36,9 +36,19 @@ Repeat=False
 
 [Common Error Pattern]
 _Name=Space before punctuation marks
-_Description=Remove space before various punctuation marks
+_Description=Add or remove space before various punctuation marks
+Classes=Human;OCR;
+# Same as the Latin pattern, except keep space before [?!;:].
+Pattern= +(["'«»]?(?!\.\.)([,.])(?!\d))
+Flags=DOTALL;MULTILINE;UNICODE;
+Replacement=\1
+Repeat=False
+Policy=Replace
+
+[Common Error Pattern]
+_Name=Space before punctuation marks
+_Description=Add or remove space before various punctuation marks
 Classes=Human;OCR;
-# Readd space removed by the corresponding pattern in Latn.
 Pattern=([^\s?!;:])([?!;:])(?!\d)
 Flags=DOTALL;MULTILINE;UNICODE;
 Replacement=\1 \2

data/patterns/Latn.common-error.in

@@ -75,7 +75,7 @@ Repeat=False
 _Name=Space before punctuation marks
 _Description=Remove space before various punctuation marks
 Classes=Human;OCR;
-# NOTE: This goes against French spacing, Latn-fr can override.
+# NOTE: Unsuitable for French, see Latn-fr.
 # Disallow trailing digits for the case of fractions, e.g. '.45'.
 Pattern= +(["'«»]?(?!\.\.)([,;:.?!])(?!\d))
 Flags=DOTALL;MULTILINE;UNICODE;
@@ -86,7 +86,8 @@ Repeat=False
 _Name=Space after punctuation marks
 _Description=Add space after various punctuation marks
 Classes=Human;OCR;
-Pattern=((\w|^|["'«»])[,;:?!])(?!["'«»])([^\W\d][\w\s])
+# NOTE: Unsuitable for Finnish, see Latn-fi.
+Pattern=((\w|^|["'«»]) ?[,;:?!])(?!["'«»])([^\W\d][\w\s])
 Flags=DOTALL;MULTILINE;UNICODE;
 Replacement=\1 \3
 Repeat=False

gaupol/patternman.py

@@ -70,18 +70,22 @@ def _filter_patterns(self, patterns):
         Patterns with a more specific code replace those with a less specific
         code if they have the same name and the more specific pattern's policy
         is explicitly set to 'Replace' (instead of the implicit 'Append').
+        Maintain order so that all patterns with the same name are always
+        located in the position of the earliest of such patterns.
         """
         filtered_patterns = []
         for i, pattern in enumerate(patterns):
-            replacement_found = False
             name = pattern.get_name(False)
-            for j in range(i + 1, len(patterns)):
-                j_name = patterns[j].get_name(False)
-                j_policy = patterns[j].get_field("Policy")
-                if (j_name == name) and (j_policy == "Replace"):
-                    replacement_found = True
-            if not replacement_found:
-                filtered_patterns.append(pattern)
+            policy = pattern.get_field("Policy")
+            last_index = len(filtered_patterns) - 1
+            for j, filtered_pattern in enumerate(filtered_patterns):
+                if filtered_pattern.get_name() == name:
+                    last_index = j
+                    if policy == "Replace":
+                        filtered_patterns[j] = None
+            filtered_patterns.insert(last_index + 1, pattern)
+            while None in filtered_patterns:
+                filtered_patterns.remove(None)
         return filtered_patterns
 
     def _get_codes_require(self, script=None, language=None, country=None):