[config/ucd] Add HB_NO_UNICODE_UNASSIGNED and activate in HB_TINY

Saves another 12kb:

$ python ./gen-ucd-table.py ucd.nounihan.grouped.zip > hb-ucd-table.hh && make -j5 CPPFLAGS='-Os -DHB_TINY' -C ~/hb/build/src/  && size ~/hb/build/src/.libs/libharfbuzz_la-hb-ucd.o
INFO: Loading UCDXML...
INFO: Preparing data tables...
INFO: Generating output...
INFO:   Compression=1:
INFO:       Dataset=gc       FullCost=18612
INFO:       Dataset=ccc      FullCost=3550
INFO:       Dataset=bmg      FullCost=1548
INFO:       Dataset=sc       FullCost=17765
INFO:       Dataset=dm       FullCost=13325
INFO:   Compression=3:
INFO:       Dataset=gc       FullCost=10726
INFO:       Dataset=ccc      FullCost=2389
INFO:       Dataset=bmg      FullCost=1052
INFO:       Dataset=sc       FullCost=13669
INFO:       Dataset=dm       FullCost=7780
INFO:   Compression=5:
INFO:       Dataset=gc       FullCost=8274
INFO:       Dataset=ccc      FullCost=2055
INFO:       Dataset=bmg      FullCost=908
INFO:       Dataset=sc       FullCost=4073
INFO:       Dataset=dm       FullCost=7780
INFO: Done.

Part of https://github.com/harfbuzz/harfbuzz/issues/1652

CONFIG.md

@@ -109,7 +109,8 @@ The pre-defined configurations are:
     as esoteric or rarely-used shaping features.  See the definition for details.
 
   * `HB_TINY`: Enables both `HB_MINI` and `HB_LEAN` configurations, as well as
-    disabling thread-safety and debugging, and use size-optimized data tables.
+    disabling thread-safety and debugging, and use even more size-optimized data
+    tables.
 
 
 ## Tailoring configuration

src/gen-ucd-table.py

@@ -72,6 +72,7 @@ for line in open('hb-common.h'):
 
 DEFAULT = 1
 COMPACT = 3
+SLOPPY  = 5
 
 
 logging.info('Generating output...')
@@ -105,15 +106,32 @@ datasets = [
     ('dm', dm, None, dm_order),
 ]
 
-for compression in (DEFAULT, COMPACT):
+for compression in (DEFAULT, COMPACT, SLOPPY):
     logging.info('  Compression=%d:' % compression)
     print()
     if compression == DEFAULT:
         print('#ifndef HB_OPTIMIZE_SIZE')
+    elif compression == COMPACT:
+        print('#elif !defined(HB_NO_UCD_UNASSIGNED)')
     else:
         print('#else')
     print()
 
+    if compression == SLOPPY:
+        for i in range(len(gc)):
+            if (i % 128) and gc[i] == 'Cn':
+                gc[i] = gc[i - 1]
+        for i in range(len(gc) - 2, -1, -1):
+            if ((i + 1) % 128) and gc[i] == 'Cn':
+                gc[i] = gc[i + 1]
+        for i in range(len(sc)):
+            if (i % 128) and sc[i] == 'Zzzz':
+                sc[i] = sc[i - 1]
+        for i in range(len(sc) - 2, -1, -1):
+            if ((i + 1) % 128) and sc[i] == 'Zzzz':
+                sc[i] = sc[i + 1]
+
+
     code = packTab.Code('_hb_ucd')
 
     for name,data,default,mapping in datasets:
@@ -123,11 +141,10 @@ for compression in (DEFAULT, COMPACT):
 
     code.print_c(linkage='static inline')
 
+    print()
 
-    if compression != DEFAULT:
+print('#endif')
-        print()
+print()
-        print('#endif')
-    print()
 
 print()
 print("#endif /* HB_UCD_TABLE_HH */")

src/hb-config.hh

@@ -40,6 +40,7 @@
 #define HB_LEAN
 #define HB_MINI
 #define HB_NO_MT
+#define HB_NO_UCD_UNASSIGNED
 #ifndef NDEBUG
 #define NDEBUG
 #endif