tdf#145925: Add DOI recognition

Detect DOI string in the form of "doi:10.*" and add hyperlink to it.
It works the same way as url recognition.

Change-Id: I3c4e78a110fd81ad7e727d5e9acee7e51127466a
Reviewed-on: https://gerrit.libreoffice.org/c/core/+/150954
Tested-by: Jenkins
Reviewed-by: Heiko Tietze <heiko.tietze@documentfoundation.org>
Reviewed-by: Stephan Bergmann <sbergman@redhat.com>

2 files changed, 113 insertions, 0 deletions

diff --git a/svl/qa/unit/test_URIHelper.cxx b/svl/qa/unit/test_URIHelper.cxx
index eb5135cbe3c6..df9e5d5114b8 100644
--- a/svl/qa/unit/test_URIHelper.cxx
+++ b/svl/qa/unit/test_URIHelper.cxx
@@ -181,11 +181,14 @@ public:
 
     void testFindFirstURLInText();
 
+    void testFindFirstDOIInText();
+
     void testResolveIdnaHost();
 
     CPPUNIT_TEST_SUITE(Test);
     CPPUNIT_TEST(testNormalizedMakeRelative);
     CPPUNIT_TEST(testFindFirstURLInText);
+    CPPUNIT_TEST(testFindFirstDOIInText);
     CPPUNIT_TEST(testResolveIdnaHost);
     CPPUNIT_TEST(finish);
     CPPUNIT_TEST_SUITE_END();
@@ -398,6 +401,57 @@ void Test::testFindFirstURLInText() {
     }
 }
 
+void Test::testFindFirstDOIInText() {
+    struct Data {
+        char const * input;
+        char const * result;
+        sal_Int32 begin;
+        sal_Int32 end;
+    };
+    static Data const tests[] = {
+        { "doi:10.1000/182", "https://doi.org/10.1000/182", 0, 15 }, // valid doi suffix with only digits
+        { "doi:10.1038/nature03001", "https://doi.org/10.1038/nature03001", 0, 23 }, // valid doi suffix with alphanumeric characters
+        { "doi:10.1093/ajae/aaq063", "https://doi.org/10.1093/ajae/aaq063", 0, 23 }, // valid doi suffix with multiple slash
+        { "doi:10.1016/S0735-1097(98)00347-7", "https://doi.org/10.1016/S0735-1097(98)00347-7", 0, 33 }, // valid doi suffix with characters apart from alphanumeric
+        { "doi:10.109/ajae/aaq063", nullptr, 0, 0 }, // # of digits after doi;10. is not between 4 and 9
+        { "doi:10.1234567890/ajae/aaq063", nullptr, 0, 0 }, // # of digits after doi;10. is not between 4 and 9
+        { "doi:10.1093/ajae/aaq063/", nullptr, 0, 0 }, // nothing after slash
+        { "doi:10.1093", nullptr, 0, 0 }, // no slash
+        { "doi:11.1093/ajae/aaq063", nullptr, 0, 0 }, // doesn't begin with doi:10.
+    };
+    CharClass charClass( m_context, LanguageTag( css::lang::Locale("en", "US", "")));
+    for (std::size_t i = 0; i < SAL_N_ELEMENTS(tests); ++i) {
+        OUString input(OUString::createFromAscii(tests[i].input));
+        sal_Int32 begin = 0;
+        sal_Int32 end = input.getLength();
+        OUString result(
+            URIHelper::FindFirstDOIInText(input, begin, end, charClass));
+        bool ok = tests[i].result == nullptr
+            ? (result.getLength() == 0 && begin == input.getLength()
+               && end == input.getLength())
+            : (result.equalsAscii(tests[i].result) && begin == tests[i].begin
+               && end == tests[i].end);
+        OString msg;
+        if (!ok) {
+            OStringBuffer buf;
+            buf.append(OString::Concat("\"")
+                + tests[i].input
+                + "\" -> ");
+            buf.append(tests[i].result == nullptr ? "none" : tests[i].result);
+            buf.append(" ("
+                + OString::number(tests[i].begin)
+                + ", "
+                + OString::number(tests[i].end)
+                + ")"
+                " != "
+                + OUStringToOString(result, RTL_TEXTENCODING_UTF8)
+                + " (" + OString::number(begin) + ", " + OString::number(end) +")");
+            msg = buf.makeStringAndClear();
+        }
+        CPPUNIT_ASSERT_MESSAGE(msg.getStr(), ok);
+    }
+}
+
 void Test::testResolveIdnaHost() {
     OUString input;
 
diff --git a/svl/source/misc/urihelper.cxx b/svl/source/misc/urihelper.cxx
index 6f121fba56d9..0043b7883a87 100644
--- a/svl/source/misc/urihelper.cxx
+++ b/svl/source/misc/urihelper.cxx
@@ -745,6 +745,65 @@ OUString URIHelper::FindFirstURLInText(OUString const & rText,
     return OUString();
 }
 
+OUString URIHelper::FindFirstDOIInText(OUString const & rText,
+                                       sal_Int32 & rBegin,
+                                       sal_Int32 & rEnd,
+                                       CharClass const & rCharClass)
+{
+    if (rBegin > rEnd || rEnd > rText.getLength())
+        return OUString();
+
+    sal_Int32 start = 7;
+    sal_Int32 count = rEnd-rBegin;
+    OUString candidate(rText.subView(rBegin, count));
+    // Match with regex "doi:10\.\d{4,9}\/[-._;()\/:a-zA-Z0-9]+"
+    if (candidate.startsWith("doi:10."))
+    {
+        bool flag = true;
+        sal_Int32 digit = 0;
+        for (sal_Int32 i=start; i<count; i++)
+        {
+            sal_Unicode c = candidate[i];
+            // Match 4 to 9 digits before slash
+            if (digit >= 0)
+            {
+                if (digit>9)
+                {
+                    flag = false;
+                    break;
+                }
+
+                if ( rCharClass.isDigit(candidate,i) )
+                {
+                    digit++;
+                }
+                else if (c=='/' && digit>=4 && i<count-1)
+                {
+                    digit=-1;
+                }
+                else
+                {
+                    flag = false;
+                    break;
+                }
+            }
+            // Match [-._;()\/:a-zA-Z0-9] after slash
+            else if (!( rCharClass.isAlphaNumeric(candidate, i) || c == '.' || c == '-' || c=='_' ||
+                        c==';' || c=='(' || c==')' || c=='\\' || (c=='/' && i<count-1) || c==':'))
+            {
+                flag = false;
+                break;
+            }
+        }
+        if (flag && digit==-1)
+        {
+            return candidate.replaceFirst("doi:","https://doi.org/");
+        }
+    }
+    rBegin = rEnd;
+    return OUString();
+}
+
 OUString URIHelper::removePassword(OUString const & rURI,
                                    INetURLObject::EncodeMechanism eEncodeMechanism,
                                    INetURLObject::DecodeMechanism eDecodeMechanism,