1/*
2*******************************************************************************
3*
4*   Copyright (C) 2000-2008, International Business Machines
5*   Corporation and others.  All Rights Reserved.
6*
7*******************************************************************************
8*   file name:  genmbcs.h
9*   encoding:   US-ASCII
10*   tab size:   8 (not used)
11*   indentation:4
12*
13*   created on: 2000jul10
14*   created by: Markus W. Scherer
15*/
16
17#ifndef __GENMBCS_H__
18#define __GENMBCS_H__
19
20#include "makeconv.h"
21
22enum {
23    /*
24     * TODO: Consider using ucnvmbcs.h constants.
25     * However, not all values need to be exactly the same, for example
26     * the xxx_UTF8_MAX values may be different. (Especially SBCS_UTF8_MAX
27     * may be higher in makeconv than in the runtime code because that
28     * affects only a small number of .cnv files [if any] but all
29     * runtime UConverterSharedData objects.
30     */
31    MBCS_STAGE_2_SHIFT=4,
32    MBCS_STAGE_2_BLOCK_SIZE=0x40,       /* =64=1<<6 for 6 bits in stage 2 */
33    MBCS_STAGE_2_BLOCK_SIZE_SHIFT=6,    /* log2(MBCS_STAGE_2_BLOCK_SIZE) */
34    MBCS_STAGE_2_BLOCK_MASK=0x3f,       /* for after shifting by MBCS_STAGE_2_SHIFT */
35    MBCS_STAGE_1_SHIFT=10,
36    MBCS_STAGE_1_BMP_SIZE=0x40, /* 0x10000>>MBCS_STAGE_1_SHIFT, or 16 for one entry per 1k code points on the BMP */
37    MBCS_STAGE_1_SIZE=0x440,    /* 0x110000>>MBCS_STAGE_1_SHIFT, or 17*64 for one entry per 1k code points */
38    MBCS_STAGE_2_SIZE=0xfbc0,   /* 0x10000-MBCS_STAGE_1_SIZE: stages 1 & 2 share a 16-bit-indexed array */
39    MBCS_MAX_STAGE_2_TOP=MBCS_STAGE_2_SIZE,
40    MBCS_STAGE_2_MAX_BLOCKS=MBCS_STAGE_2_SIZE>>MBCS_STAGE_2_BLOCK_SIZE_SHIFT,
41
42    MBCS_STAGE_2_ALL_UNASSIGNED_INDEX=0, /* stage 1 entry for the all-unassigned stage 2 block */
43    MBCS_STAGE_2_FIRST_ASSIGNED=MBCS_STAGE_2_BLOCK_SIZE, /* start of the first stage 2 block after the all-unassigned one */
44
45    MBCS_STAGE_3_BLOCK_SIZE=16,         /* =16=1<<4 for 4 bits in stage 3 */
46    MBCS_STAGE_3_BLOCK_MASK=0xf,
47    MBCS_STAGE_3_FIRST_ASSIGNED=MBCS_STAGE_3_BLOCK_SIZE, /* start of the first stage 3 block after the all-unassigned one */
48
49    MBCS_STAGE_3_GRANULARITY=16,        /* =1<<4: MBCS stage 2 indexes are shifted left 4 */
50    MBCS_STAGE_3_SBCS_SIZE=0x10000,     /* max 64k mappings for SBCS */
51    MBCS_STAGE_3_MBCS_SIZE=0x10000*MBCS_STAGE_3_GRANULARITY, /* max mappings for MBCS */
52
53    /*
54     * SBCS_UTF8_MAX: Maximum code point with UTF-8-friendly SBCS data structures.
55     * Possible values are 0x01ff..0xffff, in steps of 0x100.
56     *
57     * Unlike for MBCS, this constant only affects the stage 3 block allocation size;
58     * there is no additional stage 1/2 table stored in the .cnv file.
59     * The max value should be at least 0x7ff to cover 2-byte UTF-8.
60     * 0xfff also covers a number other small scripts which have legacy charsets
61     * (like Thai).
62     * Higher values up to 0x1fff are harmless and potentially useful because
63     * that covers small-script blocks which usually have either dense mappings
64     * or no mappings at all.
65     * Starting at U+2000, there are mostly symbols and format characters
66     * with a low density of SBCS mappings, which would result in more wasted
67     * stage 3 entries with the larger block size.
68     */
69    SBCS_UTF8_MAX=0x1fff,
70
71    /*
72     * MBCS_UTF8_MAX: Maximum code point with UTF-8-friendly MBCS data structures.
73     * Possible values are 0x01ff..0xffff, in steps of 0x100.
74     *
75     * Note that with 0xffff, MBCSAddFromUnicode() may overflow the additional UTF-8 stage table
76     * with extreme input data. The function checks for this overflow.
77     *
78     * 0xd7ff is chosen for the majority of common characters including Unihan and Hangul.
79     * At U+d800 there are mostly surrogates, private use codes, compatibility characters, etc.
80     * Larger values cause slightly larger MBCS .cnv files.
81     */
82    MBCS_UTF8_MAX=0xd7ff,
83    MBCS_UTF8_LIMIT=MBCS_UTF8_MAX+1,    /* =0xd800 */
84
85    MBCS_UTF8_STAGE_SHIFT=6,
86    MBCS_UTF8_STAGE_3_BLOCK_SIZE=0x40,  /* =64=1<<6 for 6 bits from last trail byte */
87    MBCS_UTF8_STAGE_3_BLOCK_MASK=0x3f,
88
89    /* size of the single-stage table for up to U+d7ff (used instead of stage1/2) */
90    MBCS_UTF8_STAGE_SIZE=MBCS_UTF8_LIMIT>>MBCS_UTF8_STAGE_SHIFT, /* =0x360 */
91
92    MBCS_FROM_U_EXT_FLAG=0x10,          /* UCMapping.f bit for base table mappings that fit into the base toU table */
93    MBCS_FROM_U_EXT_MASK=0x0f,          /* but need to go into the extension fromU table */
94
95    /* =4 number of regular stage 3 blocks for final UTF-8 trail byte */
96    MBCS_UTF8_STAGE_3_BLOCKS=MBCS_UTF8_STAGE_3_BLOCK_SIZE/MBCS_STAGE_3_BLOCK_SIZE,
97
98    MBCS_MAX_FALLBACK_COUNT=8192
99};
100
101U_CFUNC NewConverter *
102MBCSOpen(UCMFile *ucm);
103
104struct MBCSData;
105typedef struct MBCSData MBCSData;
106
107/*
108 * Get a dummy MBCSData for use with MBCSOkForBaseFromUnicode()
109 * for creating an extension-only file.
110 * Assume maxCharLength>1.
111 */
112U_CFUNC const MBCSData *
113MBCSGetDummy(void);
114
115/* Test if a 1:1 mapping fits into the MBCS base table's fromUnicode structure. */
116U_CFUNC UBool
117MBCSOkForBaseFromUnicode(const MBCSData *mbcsData,
118                         const uint8_t *bytes, int32_t length,
119                         UChar32 c, int8_t flag);
120
121U_CFUNC NewConverter *
122CnvExtOpen(UCMFile *ucm);
123
124#endif /* __GENMBCS_H__ */
125