src/java.desktop/share/classes/sun/font/SunLayoutEngine.java

/*
 * DO NOT ALTER OR REMOVE COPYRIGHT NOTICES OR THIS FILE HEADER.
 *
 * This code is free software; you can redistribute it and/or modify it
 * under the terms of the GNU General Public License version 2 only, as
 * published by the Free Software Foundation.  Oracle designates this
 * particular file as subject to the "Classpath" exception as provided
 * by Oracle in the LICENSE file that accompanied this code.
 *
 * This code is distributed in the hope that it will be useful, but WITHOUT
 * ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or
 * FITNESS FOR A PARTICULAR PURPOSE.  See the GNU General Public License
 * version 2 for more details (a copy is included in the LICENSE file that
 * accompanied this code).
 *
 * You should have received a copy of the GNU General Public License version
 * 2 along with this work; if not, write to the Free Software Foundation,
 * Inc., 51 Franklin St, Fifth Floor, Boston, MA 02110-1301 USA.
 *
 * Please contact Oracle, 500 Oracle Parkway, Redwood Shores, CA 94065 USA
 * or visit www.oracle.com if you need additional information or have any
 * questions.
 *
 */

/*
 *
 * (C) Copyright IBM Corp. 2003 - All Rights Reserved
 */

package sun.font;

import sun.font.GlyphLayout.*;
import sun.java2d.Disposer;
import sun.java2d.DisposerRecord;

import java.awt.geom.Point2D;
import java.lang.foreign.MemorySegment;
import java.lang.ref.SoftReference;
import java.util.concurrent.ConcurrentHashMap;
import java.util.WeakHashMap;

/*
 * different ways to do this
 * 1) each physical font2d keeps a hashtable mapping scripts to layout
 * engines, we query and fill this cache.
 * 2) we keep a mapping independent of font using the key Most likely
 * few fonts will be used, so option 2 seems better
 *
 * Once we know which engine to use for a font, we always know, so we
 * shouldn't have to recheck each time we do layout.  So the cache is
 * ok.
 *
 * Should we reuse engines?  We could instantiate an engine for each
 * font/script pair.  The engine would hold onto the table(s) from the
 * font that it needs.  If we have multiple threads using the same
 * engine, we still need to keep the state separate, so the native
 * engines would still need to be allocated for each call, since they
 * keep their state in themselves.  If they used the passed-in GVData
 * arrays directly (with some checks for space) then since each GVData
 * is different per thread, we could reuse the layout engines.  This
 * still requires a separate layout engine per font, because of the
 * table state in the engine.  If we pushed that out too and passed it
 * in with the native call as well, we'd be ok if the layout engines
 * keep all their process state on the stack, but I don't know if this
 * is true.  Then we'd basically just be down to an engine index which
 * we pass into native and then invoke the engine code (now a
 * procedure call, not an object invocation) based on a switch on the
 * index.  There would be only half a dozen engine objects then, not
 * potentially half a dozen per font.  But we'd have to stack-allocate
 * some state that included the pointer to the required font tables.
 *
 * Seems for now that the way to do things is to come in with a
 * selector and the font.  The selector indicates which engine to use,
 * the engine is stack allocated and initialized with the required
 * font tables (the selector indicates which).  Then layout is called,
 * the contents are copied (or not), and the stack is destroyed on
 * exit. So the association is between the font/script (layout engine
 * desc) and one of a few permanent engine objects, which are
 * handed the key when they need to process something.  In the native
 * case, the engine holds an index, and just passes it together with
 * the key info down to native.  Some default cases are the 'default
 * layout' case that just runs the c2gmapper, this stays in java and
 * just uses the mapper from the font/strike.  Another default case
 * might be the unicode arabic shaper, since this doesn't care about
 * the font (or script or lang?) it wouldn't need to extract this
 * data.  It could be (yikes) ported back to java even to avoid
 * upcalls to check if the font supports a particular unicode
 * character.
 *
 * I'd expect that the majority of scripts use the default mapper for
 * a particular font.  Loading the hashtable with 40 or so keys 30+ of
 * which all map to the same object is unfortunate.  It might be worth
 * instead having a per-font list of 'scripts with non-default
 * engines', e.g. the factory has a hashtable mapping fonts to 'script
 * lists' (the factory has this since the design potentially has other
 * factories, though I admit there's no client for this yet and no
 * public api) and then the script list is queried for the script in
 * question.  it can be preloaded at creation time with all the
 * scripts that don't have default engines-- either a list or a hash
 * table, so a null return from the table means 'default' and not 'i
 * don't know yet'.
 *
 * On the other hand, in most all cases the number of unique
 * script/font combinations will be small, so a flat hashtable should
 * suffice.
 * */
public final class SunLayoutEngine implements LayoutEngine, LayoutEngineFactory {
    static {
        FontManagerNativeLibrary.load();
    }

    private LayoutEngineKey key;

    private static LayoutEngineFactory instance;

    public static LayoutEngineFactory instance() {
        if (instance == null) {
            instance = new SunLayoutEngine();
        }
        return instance;
    }

    private SunLayoutEngine() {
        // actually a factory, key is null so layout cannot be called on it
    }

    public LayoutEngine getEngine(Font2D font, int script, int lang) {
        return getEngine(new LayoutEngineKey(font, script, lang));
    }

  // !!! don't need this unless we have more than one sun layout engine...
    public LayoutEngine getEngine(LayoutEngineKey key) {
        ConcurrentHashMap<LayoutEngineKey, LayoutEngine> cache = cacheref.get();
        if (cache == null) {
            cache = new ConcurrentHashMap<>();
            cacheref = new SoftReference<>(cache);
        }

        LayoutEngine e = cache.get(key);
        if (e == null) {
            LayoutEngineKey copy = key.copy();
            e = new SunLayoutEngine(copy);
            cache.put(copy, e);
        }
        return e;
    }
    private SoftReference<ConcurrentHashMap<LayoutEngineKey, LayoutEngine>> cacheref =
        new SoftReference<>(null);

    private SunLayoutEngine(LayoutEngineKey key) {
        this.key = key;
    }

    private static final WeakHashMap<Font2D, FaceRef> facePtr =
            new WeakHashMap<>();

    private long getFacePtr(Font2D font2D) {
        FaceRef ref;
        synchronized (facePtr) {
            ref = facePtr.computeIfAbsent(font2D, FaceRef::new);
        }
        return ref.getNativePtr();
    }

    static boolean useFFM = true;
    static {
        String prop = System.getProperty("sun.font.layout.ffm", "true");
        useFFM = "true".equals(prop);

    }

    public void layout(FontStrikeDesc desc, float[] mat, float ptSize, int gmask,
                       int baseIndex, TextRecord tr, int typo_flags,
                       Point2D.Float pt, GVData data) {

        Font2D font = key.font();
        FontStrike strike = font.getStrike(desc);
        if (useFFM) {
            MemorySegment face = HBShaper.getFace(font);
            if (face != null) {
                HBShaper.shape(font, strike, ptSize, mat, face,
                        tr.text, data, key.script(),
                        tr.start, tr.limit, baseIndex, pt,
                        typo_flags, gmask);
            }
        } else {
            long pFace = getFacePtr(font);
            if (pFace != 0) {
                shape(font, strike, ptSize, mat, pFace,
                    tr.text, data, key.script(),
                    tr.start, tr.limit, baseIndex, pt,
                    typo_flags, gmask);
            }
        }
    }

    /* Native method to invoke harfbuzz layout engine */
    private static native boolean
        shape(Font2D font, FontStrike strike, float ptSize, float[] mat,
              long pFace,
              char[] chars, GVData data,
              int script, int offset, int limit,
              int baseIndex, Point2D.Float pt, int typo_flags, int slot);

    private static native long createFace(Font2D font,
                                          long platformNativeFontPtr);

    private static native void disposeFace(long facePtr);

    private static class FaceRef implements DisposerRecord {
        private Font2D font;
        private Long facePtr;

        private FaceRef(Font2D font) {
            this.font = font;
        }

        private synchronized long getNativePtr() {
            if (facePtr == null) {
                facePtr = createFace(font,
                        font.getPlatformNativeFontPtr());
                if (facePtr != 0) {
                    Disposer.addObjectRecord(font, this);
                }
                font = null;
            }
            return facePtr;
        }

        @Override
        public void dispose() {
            disposeFace(facePtr);
        }
    }
}