ICU-11443 WIP: Link detection according to UTS#58.

icu4c/source/common/linkemailprops.h

@@ -0,0 +1,45 @@
+// © 2025 and later: Unicode, Inc. and others.
+// License & terms of use: https://www.unicode.org/copyright.html
+
+// linkemailprops.h
+// created: 2025 for UTS #58 / Unicode 17.0
+
+#ifndef LINKEMAILPROPS_H
+#define LINKEMAILPROPS_H
+
+#include "unicode/utypes.h"
+#include "unicode/ucptrie.h"
+#include "unicode/uobject.h"
+
+U_NAMESPACE_BEGIN
+
+/**
+ * Link_Email binary property constants and data-file identifiers.
+ * A code point has Link_Email=Yes (1) if it is allowed in an email local part.
+ * All other code points default to No (0).
+ */
+class LinkEmailProps : public UMemory {
+public:
+    /**
+     * Indexes into the binary data indexes[] array.
+     * Values are byte offsets from the start of the indexes[] array.
+     */
+    enum {
+        IX_COUNT,       // 0: length of indexes[] (== IX_LINK_EMAIL_COUNT)
+        IX_CPTRIE_TOP,  // 1: limit offset of the Link_Email UCPTrie
+        IX_TRIE2_TOP,   // 2: reserved for a second future trie (= IX_CPTRIE_TOP until used)
+        IX_TRIE3_TOP,   // 3: reserved for a third future trie  (= IX_TRIE2_TOP  until used)
+        IX_TOTAL_SIZE,  // 4: total data size (= limit of last trie)
+        // reserved
+        IX_LINK_EMAIL_COUNT = 8
+    };
+
+    static constexpr char DATA_TYPE[] = "icu";
+    static constexpr char DATA_NAME[] = "ulinkemail";
+    static constexpr uint8_t DATA_FORMAT[4] = { 'L', 'n', 'k', 'E' };
+    static constexpr uint8_t FORMAT_VERSION[4] = { 1, 0, 0, 0 };
+};
+
+U_NAMESPACE_END
+
+#endif  // LINKEMAILPROPS_H

icu4c/source/common/linktermprops.h

@@ -0,0 +1,53 @@
+// © 2025 and later: Unicode, Inc. and others.
+// License & terms of use: https://www.unicode.org/copyright.html
+
+// linktermprops.h
+// created: 2025 for UTS #58 / Unicode 17.0
+
+#ifndef LINKTERMPROPS_H
+#define LINKTERMPROPS_H
+
+#include "unicode/utypes.h"
+#include "unicode/ucptrie.h"
+#include "unicode/uobject.h"
+
+/**
+ * Values of the Link_Term property (UTS #58 / proposed Unicode 19.0).
+ * The default value for unlisted code points is ULINK_TERM_HARD.
+ */
+typedef enum ULinkTerm {
+    ULINK_TERM_HARD    = 0,  /**< Terminates a URL unconditionally. Default. */
+    ULINK_TERM_INCLUDE = 1,  /**< May appear in a URL (letters, digits, …). */
+    ULINK_TERM_SOFT    = 2,  /**< Terminates only when followed by Hard. */
+    ULINK_TERM_CLOSE   = 3,  /**< Closing bracket; terminates if unmatched. */
+    ULINK_TERM_OPEN    = 4,  /**< Opening bracket. */
+    ULINK_TERM_COUNT
+} ULinkTerm;
+
+U_NAMESPACE_BEGIN
+
+class LinkTermProps : public UMemory {
+public:
+    /**
+     * Indexes into the binary data indexes[] array.
+     * Values are byte offsets from the start of the indexes[] array.
+     */
+    enum {
+        IX_COUNT,       // 0: length of indexes[] (== IX_LINK_TERM_COUNT)
+        IX_CPTRIE_TOP,  // 1: limit offset of the Link_Term UCPTrie
+        IX_TRIE2_TOP,   // 2: reserved for a second future trie (= IX_CPTRIE_TOP until used)
+        IX_TRIE3_TOP,   // 3: reserved for a third future trie  (= IX_TRIE2_TOP  until used)
+        IX_TOTAL_SIZE,  // 4: total data size (= limit of last trie)
+        // reserved
+        IX_LINK_TERM_COUNT = 8
+    };
+
+    static constexpr char DATA_TYPE[] = "icu";
+    static constexpr char DATA_NAME[] = "ulinkterm";
+    static constexpr uint8_t DATA_FORMAT[4] = { 'L', 'n', 'k', 'T' };
+    static constexpr uint8_t FORMAT_VERSION[4] = { 1, 0, 0, 0 };
+};
+
+U_NAMESPACE_END
+
+#endif  // LINKTERMPROPS_H

icu4j/main/core/pom.xml

@@ -45,6 +45,23 @@
 
   <build>
     <plugins>
+      <plugin>
+        <groupId>org.codehaus.mojo</groupId>
+        <artifactId>exec-maven-plugin</artifactId>
+        <executions>
+          <execution>
+            <id>generate-iana-tlds</id>
+            <phase>generate-sources</phase>
+            <goals><goal>exec</goal></goals>
+            <configuration>
+              <executable>python3</executable>
+              <arguments>
+                <argument>${project.basedir}/src/main/scripts/generate-iana-tlds.py</argument>
+              </arguments>
+            </configuration>
+          </execution>
+        </executions>
+      </plugin>
       <plugin>
         <artifactId>maven-jar-plugin</artifactId>
         <executions>

icu4j/main/core/src/main/java/com/ibm/icu/impl/LinkEmailProps.java

@@ -0,0 +1,71 @@
+// © 2025 and later: Unicode, Inc. and others.
+// License & terms of use: https://www.unicode.org/copyright.html
+
+package com.ibm.icu.impl;
+
+import com.ibm.icu.util.CodePointTrie;
+import com.ibm.icu.util.ICUUncheckedIOException;
+import java.io.IOException;
+import java.nio.ByteBuffer;
+
+/**
+ * Link_Email binary property loaded from ulinkemail.icu.
+ * Implements the Link_Email property (UTS #58 / Unicode 17.0).
+ *
+ * <p>A code point has Link_Email=Yes if it may appear in an email local part.
+ * All other code points have Link_Email=No (the default, stored as 0).
+ */
+public final class LinkEmailProps {
+
+    // Indexes into the binary data indexes[] array (see linkemailprops.h).
+    private static final int IX_COUNT      = 0;
+    private static final int IX_CPTRIE_TOP = 1;
+
+    // "LnkE"
+    private static final int DATA_FORMAT = 0x4C6E6B45;
+
+    private static final ICUBinary.Authenticate IS_ACCEPTABLE =
+            version -> version[0] == 1;
+
+    public static final LinkEmailProps INSTANCE = new LinkEmailProps();
+
+    private final CodePointTrie.Fast8 cpTrie;
+
+    private LinkEmailProps() {
+        ByteBuffer bytes = ICUBinary.getRequiredData("ulinkemail.icu");
+        try {
+            ICUBinary.readHeaderAndDataVersion(bytes, DATA_FORMAT, IS_ACCEPTABLE);
+            int startPos = bytes.position();
+
+            // indexes[0] = number of entries in the indexes array.
+            int indexCount = bytes.getInt();
+            if (indexCount < 2) {
+                throw new ICUUncheckedIOException("ulinkemail.icu: indexes too short");
+            }
+            int[] inIndexes = new int[indexCount];
+            inIndexes[IX_COUNT] = indexCount;
+            for (int i = 1; i < indexCount; i++) {
+                inIndexes[i] = bytes.getInt();
+            }
+
+            // The UCPTrie starts immediately after the indexes[] array and
+            // ends at inIndexes[IX_CPTRIE_TOP] (a byte offset from startPos).
+            cpTrie = CodePointTrie.Fast8.fromBinary(bytes);
+            int pos = bytes.position() - startPos;
+            ICUBinary.skipBytes(bytes, inIndexes[IX_CPTRIE_TOP] - pos);
+        } catch (IOException e) {
+            throw new ICUUncheckedIOException(e);
+        }
+    }
+
+    /**
+     * Returns true if the code point has Link_Email=Yes,
+     * i.e., it is allowed in an email local part.
+     *
+     * @param c a Unicode code point
+     * @return true if {@code c} has Link_Email=Yes
+     */
+    public boolean contains(int c) {
+        return cpTrie.get(c) != 0;
+    }
+}

icu4j/main/core/src/main/java/com/ibm/icu/impl/LinkTermProps.java

@@ -0,0 +1,84 @@
+// © 2025 and later: Unicode, Inc. and others.
+// License & terms of use: https://www.unicode.org/copyright.html
+
+package com.ibm.icu.impl;
+
+import com.ibm.icu.util.CodePointTrie;
+import com.ibm.icu.util.ICUUncheckedIOException;
+import java.io.IOException;
+import java.nio.ByteBuffer;
+
+/**
+ * Link termination properties loaded from ulinkterm.icu.
+ * Implements the Link_Term property (UTS #58 / proposed Unicode 19.0).
+ *
+ * <p>Values match the C-side ULinkTerm enum:
+ * <ul>
+ *   <li>{@link #HARD}    = 0 (default for all unlisted code points)</li>
+ *   <li>{@link #INCLUDE} = 1</li>
+ *   <li>{@link #SOFT}    = 2</li>
+ *   <li>{@link #CLOSE}   = 3</li>
+ *   <li>{@link #OPEN}    = 4</li>
+ * </ul>
+ */
+public final class LinkTermProps {
+
+    // Link_Term property values — must match ULinkTerm in linktermprops.h.
+    public static final int HARD    = 0;
+    public static final int INCLUDE = 1;
+    public static final int SOFT    = 2;
+    public static final int CLOSE   = 3;
+    public static final int OPEN    = 4;
+
+    // Indexes into the binary data indexes[] array (see linktermprops.h).
+    private static final int IX_COUNT      = 0;
+    private static final int IX_CPTRIE_TOP = 1;
+
+    // "LnkT"
+    private static final int DATA_FORMAT = 0x4C6E6B54;
+
+    private static final ICUBinary.Authenticate IS_ACCEPTABLE =
+            version -> version[0] == 1;
+
+    public static final LinkTermProps INSTANCE = new LinkTermProps();
+
+    private final CodePointTrie.Fast8 cpTrie;
+
+    private LinkTermProps() {
+        ByteBuffer bytes = ICUBinary.getRequiredData("ulinkterm.icu");
+        try {
+            ICUBinary.readHeaderAndDataVersion(bytes, DATA_FORMAT, IS_ACCEPTABLE);
+            int startPos = bytes.position();
+
+            // indexes[0] = number of entries in the indexes array.
+            int indexCount = bytes.getInt();
+            if (indexCount < 2) {
+                throw new ICUUncheckedIOException("ulinkterm.icu: indexes too short");
+            }
+            int[] inIndexes = new int[indexCount];
+            inIndexes[IX_COUNT] = indexCount;
+            for (int i = 1; i < indexCount; i++) {
+                inIndexes[i] = bytes.getInt();
+            }
+
+            // The UCPTrie starts immediately after the indexes[] array and
+            // ends at inIndexes[IX_CPTRIE_TOP] (a byte offset from startPos).
+            cpTrie = CodePointTrie.Fast8.fromBinary(bytes);
+            int pos = bytes.position() - startPos;
+            ICUBinary.skipBytes(bytes, inIndexes[IX_CPTRIE_TOP] - pos);
+        } catch (IOException e) {
+            throw new ICUUncheckedIOException(e);
+        }
+    }
+
+    /**
+     * Returns the Link_Term value for a code point.
+     *
+     * @param c a Unicode code point
+     * @return one of {@link #HARD}, {@link #INCLUDE}, {@link #SOFT},
+     *         {@link #CLOSE}, {@link #OPEN}
+     */
+    public int get(int c) {
+        return cpTrie.get(c);
+    }
+}