Improve copyright detection

Handle more edge cases

Signed-off-by: Philippe Ombredanne <[email protected]>

Author: pombredanne

src/cluecode/copyrights.py

@@ -2335,6 +2335,9 @@ def build_detection_from_node(
     # created by Jason Hunter <jhunter AT jdom DOT org>
     EMAIL: {<EMAIL_START>  <AT>  <NN|NNP>  <DOT>  <NN|NNP> } # email_start
 
+    # Copyright (c) 2001 Karl Garrison (karl AT indy.rr.com)
+    EMAIL: {<NN|NNP>  <AT>  <URL> } # email_at
+
     EMAIL: { <NN>  <CC>  <NN>  <DOT>  <NN> } # foo at bat dot com
 
     # [email protected] or [email protected]

src/licensedcode/data/rules/jdom_2.RULE

@@ -2,7 +2,7 @@
 license_expression: jdom
 is_license_text: yes
 ignorable_authors:
-    - Jason Hunter
+    - Jason Hunter jhunter AT jdom DOT org and Brett McLaughlin brett AT jdom DOT org
     - the JDOM Project (http://www.jdom.org/)
 ignorable_urls:
     - http://www.jdom.org/
@@ -51,4 +51,4 @@ ADVISED OF THE POSSIBILITY OF SUCH DAMAGE.
 This software consists of voluntary contributions made by many individuals on
 behalf of the JDOM Project  and was originally created by Jason Hunter <jhunter
 AT jdom DOT org> and Brett McLaughlin <brett AT jdom DOT org>.  For more
-information on the JDOM Project, please see <http://www.jdom.org/>.
+information on the JDOM Project, please see <http://www.jdom.org/>.

src/textcode/markup.py

@@ -1918,7 +1918,6 @@ def strip_known_markup_from_text(text):
         "<url>",
         "</url>",
         "<date-of-document>",
-
         # common XML namespaces
         "http://www.w3.org/1998/math/mathml",
         "http://www.w3.org/1999/xhtml",
@@ -1937,7 +1936,11 @@ def strip_known_markup_from_text(text):
     "@end",
     "@group",
     "mailto:",
-    "xmlns",
+    "xmlns=",
+    "xmlns:",
+    "xml:",
+    "lang=",
+    "<windows"
 )
 
 

tests/cluecode/data/copyright_fossology/testdata118.yml

@@ -39,4 +39,6 @@ holders:
   - Guy Eric Schalnat, Group 42, Inc.
   - Jean-loup Gailly and Mark Adler
 authors:
+authors:
+  - Masayuki Hatta (mhatta)
   - artofcode LLC. http://artofcode.com

tests/cluecode/data/copyright_fossology/testdata119.yml

@@ -39,4 +39,6 @@ holders:
   - Guy Eric Schalnat, Group 42, Inc.
   - Jean-loup Gailly and Mark Adler
 authors:
+authors:
+  - Masayuki Hatta (mhatta)
   - artofcode LLC. http://artofcode.com

tests/cluecode/data/copyright_fossology/testdata127.yml

@@ -7,4 +7,5 @@ copyrights:
 holders:
   - Valek Filippov
 authors:
+  - Masayuki Hatta (mhatta)
   - URW++ Design & Development GmbH Valek Filippov <[email protected]>

tests/cluecode/data/copyright_fossology/testdata128.yml

@@ -39,4 +39,5 @@ holders:
   - Guy Eric Schalnat, Group 42, Inc.
   - Jean-loup Gailly and Mark Adler
 authors:
+  - Masayuki Hatta (mhatta)
   - artofcode LLC. http://artofcode.com

tests/cluecode/data/copyright_fossology/testdata19.yml

@@ -9,14 +9,14 @@ copyrights:
 holders:
   - Karl Garrison
   - Martin R.
-  - Karl Garrison karl AT
+  - Karl Garrison
 authors:
   - firstname Karl Garrison
   - Martin R. Jones
-  - Martin R. Jones xmlns:doc http://nwalsh.com/xsl/documentation/1.0
-  - Jean-Claude Dumas xmlns:doc http://nwalsh.com/xsl/documentation/1.0
-  - Jean-Claude surname Dumas xmlns:doc http://nwalsh.com/xsl/documentation/1.0
+  - Martin R. Jones
+  - Jean-Claude Dumas
+  - Jean-Claude surname Dumas
   - Masayuki Koba
-  - Frank Pieczynski xmlns:doc http://nwalsh.com/xsl/documentation/1.0
-  - Bartosz Trudnowski for his wife. xmlns:doc http://nwalsh.com/xsl/documentation/1.0
+  - Frank Pieczynski
+  - Bartosz Trudnowski
   - Daniel Pfeiffer

tests/cluecode/data/copyrights/abiword_common.copyright.yml

@@ -8,12 +8,12 @@ copyrights:
   - Copyright (c) 2009 Patrik Fimml <[email protected]>
 holders:
   - AbiSource, Inc. & Co.
-  - Masayuki Hatta (mhatta)
+  - Masayuki Hatta
   - Patrik Fimml
 holders_summary:
   - value: AbiSource, Inc. & Co.
     count: 1
-  - value: Masayuki Hatta (mhatta)
+  - value: Masayuki Hatta
     count: 1
   - value: Patrik Fimml
     count: 1

tests/cluecode/data/copyrights/misco4/to_improve/junk-copyright-333.txt.yml

@@ -2,7 +2,4 @@ what:
   - copyrights
   - holders
   - authors
-copyrights:
-  - (c) AT r'b
-holders:
-  - r'b
+