diff mbox series

[06/11] vt: introduce gen_ucs_recompose.py to create ucs_recompose.c

Message ID 20250410011839.64418-7-nico@fluxnic.net
State Superseded
Headers show
Series vt: implement proper Unicode handling | expand

Commit Message

Nicolas Pitre April 10, 2025, 1:13 a.m. UTC
From: Nicolas Pitre <npitre@baylibre.com>

The generated code includes a table that maps base character + combining
mark pairs to their precomposed equivalents using Python's unicodedata
module. It also provides the ucs_recompose() function to query that
table.

The default script behavior is to create a table with most commonly used
Latin, Greek, and Cyrillic recomposition pairs only. It is much smaller
than the table with all possible recomposition pairs (71 entries vs 1000
entries). But if one needs/wants the full table then simply running the
script with the --full argument will generate it.

Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
---
 drivers/tty/vt/gen_ucs_recompose.py | 321 ++++++++++++++++++++++++++++
 1 file changed, 321 insertions(+)
 create mode 100755 drivers/tty/vt/gen_ucs_recompose.py

Comments

Jiri Slaby April 14, 2025, 7:08 a.m. UTC | #1
On 10. 04. 25, 3:13, Nicolas Pitre wrote:
> From: Nicolas Pitre <npitre@baylibre.com>
> 
> The generated code includes a table that maps base character + combining
> mark pairs to their precomposed equivalents using Python's unicodedata
> module. It also provides the ucs_recompose() function to query that
> table.
> 
> The default script behavior is to create a table with most commonly used
> Latin, Greek, and Cyrillic recomposition pairs only. It is much smaller
> than the table with all possible recomposition pairs (71 entries vs 1000
> entries). But if one needs/wants the full table then simply running the
> script with the --full argument will generate it.
> 
> Signed-off-by: Nicolas Pitre <npitre@baylibre.com>
> ---
>   drivers/tty/vt/gen_ucs_recompose.py | 321 ++++++++++++++++++++++++++++
>   1 file changed, 321 insertions(+)
>   create mode 100755 drivers/tty/vt/gen_ucs_recompose.py
> 
> diff --git a/drivers/tty/vt/gen_ucs_recompose.py b/drivers/tty/vt/gen_ucs_recompose.py
> new file mode 100755
> index 0000000000..64418803e4
> --- /dev/null
> +++ b/drivers/tty/vt/gen_ucs_recompose.py
...
> +struct compare_key {{
> +	uint16_t base;
> +	uint16_t combining;
> +}};
> +
> +static int recomposition_compare(const void *key, const void *element)
> +{{
> +	const struct compare_key *search_key = key;
> +	const struct recomposition *table_entry = element;
> +
> +	/* Compare base character first */
> +	if (search_key->base < table_entry->base)
> +		return -1;
> +	if (search_key->base > table_entry->base)
> +		return 1;
> +
> +	/* Base characters match, now compare combining character */
> +	if (search_key->combining < table_entry->combining)
> +		return -1;
> +	if (search_key->combining > table_entry->combining)
> +		return 1;
> +
> +	/* Both match */
> +	return 0;
> +}}
> +
> +/**
> + * Attempt to recompose two Unicode characters into a single character.
> + *
> + * @param previous: Previous Unicode code point (UCS-4)
> + * @param current: Current Unicode code point (UCS-4)
> + * Return: Recomposed Unicode code point, or 0 if no recomposition is possible
> + */
> +uint32_t ucs_recompose(uint32_t base, uint32_t combining)
> +{{
> +	/* Check if characters are within the range of our table */
> +	if (base < MIN_BASE_CHAR || base > MAX_BASE_CHAR ||
> +	    combining < MIN_COMBINING_CHAR || combining > MAX_COMBINING_CHAR)
> +		return 0;
> +
> +	struct compare_key key = {{ base, combining }};
> +
> +	struct recomposition *result =
> +		__inline_bsearch(&key, recomposition_table,
> +				 ARRAY_SIZE(recomposition_table),
> +				 sizeof(*recomposition_table),
> +				 recomposition_compare);
> +
> +	return result ? result->recomposed : 0;
> +}}

Again, I think no reason to maintain C functions in py.

thanks,
diff mbox series

Patch

diff --git a/drivers/tty/vt/gen_ucs_recompose.py b/drivers/tty/vt/gen_ucs_recompose.py
new file mode 100755
index 0000000000..64418803e4
--- /dev/null
+++ b/drivers/tty/vt/gen_ucs_recompose.py
@@ -0,0 +1,321 @@ 
+#!/usr/bin/env python3
+# SPDX-License-Identifier: GPL-2.0
+#
+# This script uses Python's unicodedata module to generate ucs_recompose.c.
+# The generated code maps base character + combining mark pairs to their
+# precomposed equivalents.
+#
+# Usage:
+#   python gen_ucs_recompose.py         # Generate with common recomposition pairs
+#   python gen_ucs_recompose.py --full  # Generate with all recomposition pairs
+
+import unicodedata
+import sys
+import argparse
+import textwrap
+
+common_recompose_description = "most commonly used Latin, Greek, and Cyrillic recomposition pairs only"
+COMMON_RECOMPOSITION_PAIRS = [
+    # Latin letters with accents - uppercase
+    (0x0041, 0x0300, 0x00C0),  # A + COMBINING GRAVE ACCENT = LATIN CAPITAL LETTER A WITH GRAVE
+    (0x0041, 0x0301, 0x00C1),  # A + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER A WITH ACUTE
+    (0x0041, 0x0302, 0x00C2),  # A + COMBINING CIRCUMFLEX ACCENT = LATIN CAPITAL LETTER A WITH CIRCUMFLEX
+    (0x0041, 0x0303, 0x00C3),  # A + COMBINING TILDE = LATIN CAPITAL LETTER A WITH TILDE
+    (0x0041, 0x0308, 0x00C4),  # A + COMBINING DIAERESIS = LATIN CAPITAL LETTER A WITH DIAERESIS
+    (0x0041, 0x030A, 0x00C5),  # A + COMBINING RING ABOVE = LATIN CAPITAL LETTER A WITH RING ABOVE
+    (0x0043, 0x0327, 0x00C7),  # C + COMBINING CEDILLA = LATIN CAPITAL LETTER C WITH CEDILLA
+    (0x0045, 0x0300, 0x00C8),  # E + COMBINING GRAVE ACCENT = LATIN CAPITAL LETTER E WITH GRAVE
+    (0x0045, 0x0301, 0x00C9),  # E + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER E WITH ACUTE
+    (0x0045, 0x0302, 0x00CA),  # E + COMBINING CIRCUMFLEX ACCENT = LATIN CAPITAL LETTER E WITH CIRCUMFLEX
+    (0x0045, 0x0308, 0x00CB),  # E + COMBINING DIAERESIS = LATIN CAPITAL LETTER E WITH DIAERESIS
+    (0x0049, 0x0300, 0x00CC),  # I + COMBINING GRAVE ACCENT = LATIN CAPITAL LETTER I WITH GRAVE
+    (0x0049, 0x0301, 0x00CD),  # I + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER I WITH ACUTE
+    (0x0049, 0x0302, 0x00CE),  # I + COMBINING CIRCUMFLEX ACCENT = LATIN CAPITAL LETTER I WITH CIRCUMFLEX
+    (0x0049, 0x0308, 0x00CF),  # I + COMBINING DIAERESIS = LATIN CAPITAL LETTER I WITH DIAERESIS
+    (0x004E, 0x0303, 0x00D1),  # N + COMBINING TILDE = LATIN CAPITAL LETTER N WITH TILDE
+    (0x004F, 0x0300, 0x00D2),  # O + COMBINING GRAVE ACCENT = LATIN CAPITAL LETTER O WITH GRAVE
+    (0x004F, 0x0301, 0x00D3),  # O + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER O WITH ACUTE
+    (0x004F, 0x0302, 0x00D4),  # O + COMBINING CIRCUMFLEX ACCENT = LATIN CAPITAL LETTER O WITH CIRCUMFLEX
+    (0x004F, 0x0303, 0x00D5),  # O + COMBINING TILDE = LATIN CAPITAL LETTER O WITH TILDE
+    (0x004F, 0x0308, 0x00D6),  # O + COMBINING DIAERESIS = LATIN CAPITAL LETTER O WITH DIAERESIS
+    (0x0055, 0x0300, 0x00D9),  # U + COMBINING GRAVE ACCENT = LATIN CAPITAL LETTER U WITH GRAVE
+    (0x0055, 0x0301, 0x00DA),  # U + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER U WITH ACUTE
+    (0x0055, 0x0302, 0x00DB),  # U + COMBINING CIRCUMFLEX ACCENT = LATIN CAPITAL LETTER U WITH CIRCUMFLEX
+    (0x0055, 0x0308, 0x00DC),  # U + COMBINING DIAERESIS = LATIN CAPITAL LETTER U WITH DIAERESIS
+    (0x0059, 0x0301, 0x00DD),  # Y + COMBINING ACUTE ACCENT = LATIN CAPITAL LETTER Y WITH ACUTE
+
+    # Latin letters with accents - lowercase
+    (0x0061, 0x0300, 0x00E0),  # a + COMBINING GRAVE ACCENT = LATIN SMALL LETTER A WITH GRAVE
+    (0x0061, 0x0301, 0x00E1),  # a + COMBINING ACUTE ACCENT = LATIN SMALL LETTER A WITH ACUTE
+    (0x0061, 0x0302, 0x00E2),  # a + COMBINING CIRCUMFLEX ACCENT = LATIN SMALL LETTER A WITH CIRCUMFLEX
+    (0x0061, 0x0303, 0x00E3),  # a + COMBINING TILDE = LATIN SMALL LETTER A WITH TILDE
+    (0x0061, 0x0308, 0x00E4),  # a + COMBINING DIAERESIS = LATIN SMALL LETTER A WITH DIAERESIS
+    (0x0061, 0x030A, 0x00E5),  # a + COMBINING RING ABOVE = LATIN SMALL LETTER A WITH RING ABOVE
+    (0x0063, 0x0327, 0x00E7),  # c + COMBINING CEDILLA = LATIN SMALL LETTER C WITH CEDILLA
+    (0x0065, 0x0300, 0x00E8),  # e + COMBINING GRAVE ACCENT = LATIN SMALL LETTER E WITH GRAVE
+    (0x0065, 0x0301, 0x00E9),  # e + COMBINING ACUTE ACCENT = LATIN SMALL LETTER E WITH ACUTE
+    (0x0065, 0x0302, 0x00EA),  # e + COMBINING CIRCUMFLEX ACCENT = LATIN SMALL LETTER E WITH CIRCUMFLEX
+    (0x0065, 0x0308, 0x00EB),  # e + COMBINING DIAERESIS = LATIN SMALL LETTER E WITH DIAERESIS
+    (0x0069, 0x0300, 0x00EC),  # i + COMBINING GRAVE ACCENT = LATIN SMALL LETTER I WITH GRAVE
+    (0x0069, 0x0301, 0x00ED),  # i + COMBINING ACUTE ACCENT = LATIN SMALL LETTER I WITH ACUTE
+    (0x0069, 0x0302, 0x00EE),  # i + COMBINING CIRCUMFLEX ACCENT = LATIN SMALL LETTER I WITH CIRCUMFLEX
+    (0x0069, 0x0308, 0x00EF),  # i + COMBINING DIAERESIS = LATIN SMALL LETTER I WITH DIAERESIS
+    (0x006E, 0x0303, 0x00F1),  # n + COMBINING TILDE = LATIN SMALL LETTER N WITH TILDE
+    (0x006F, 0x0300, 0x00F2),  # o + COMBINING GRAVE ACCENT = LATIN SMALL LETTER O WITH GRAVE
+    (0x006F, 0x0301, 0x00F3),  # o + COMBINING ACUTE ACCENT = LATIN SMALL LETTER O WITH ACUTE
+    (0x006F, 0x0302, 0x00F4),  # o + COMBINING CIRCUMFLEX ACCENT = LATIN SMALL LETTER O WITH CIRCUMFLEX
+    (0x006F, 0x0303, 0x00F5),  # o + COMBINING TILDE = LATIN SMALL LETTER O WITH TILDE
+    (0x006F, 0x0308, 0x00F6),  # o + COMBINING DIAERESIS = LATIN SMALL LETTER O WITH DIAERESIS
+    (0x0075, 0x0300, 0x00F9),  # u + COMBINING GRAVE ACCENT = LATIN SMALL LETTER U WITH GRAVE
+    (0x0075, 0x0301, 0x00FA),  # u + COMBINING ACUTE ACCENT = LATIN SMALL LETTER U WITH ACUTE
+    (0x0075, 0x0302, 0x00FB),  # u + COMBINING CIRCUMFLEX ACCENT = LATIN SMALL LETTER U WITH CIRCUMFLEX
+    (0x0075, 0x0308, 0x00FC),  # u + COMBINING DIAERESIS = LATIN SMALL LETTER U WITH DIAERESIS
+    (0x0079, 0x0301, 0x00FD),  # y + COMBINING ACUTE ACCENT = LATIN SMALL LETTER Y WITH ACUTE
+    (0x0079, 0x0308, 0x00FF),  # y + COMBINING DIAERESIS = LATIN SMALL LETTER Y WITH DIAERESIS
+
+    # Common Greek characters
+    (0x0391, 0x0301, 0x0386),  # Α + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER ALPHA WITH TONOS
+    (0x0395, 0x0301, 0x0388),  # Ε + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER EPSILON WITH TONOS
+    (0x0397, 0x0301, 0x0389),  # Η + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER ETA WITH TONOS
+    (0x0399, 0x0301, 0x038A),  # Ι + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER IOTA WITH TONOS
+    (0x039F, 0x0301, 0x038C),  # Ο + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER OMICRON WITH TONOS
+    (0x03A5, 0x0301, 0x038E),  # Υ + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER UPSILON WITH TONOS
+    (0x03A9, 0x0301, 0x038F),  # Ω + COMBINING ACUTE ACCENT = GREEK CAPITAL LETTER OMEGA WITH TONOS
+    (0x03B1, 0x0301, 0x03AC),  # α + COMBINING ACUTE ACCENT = GREEK SMALL LETTER ALPHA WITH TONOS
+    (0x03B5, 0x0301, 0x03AD),  # ε + COMBINING ACUTE ACCENT = GREEK SMALL LETTER EPSILON WITH TONOS
+    (0x03B7, 0x0301, 0x03AE),  # η + COMBINING ACUTE ACCENT = GREEK SMALL LETTER ETA WITH TONOS
+    (0x03B9, 0x0301, 0x03AF),  # ι + COMBINING ACUTE ACCENT = GREEK SMALL LETTER IOTA WITH TONOS
+    (0x03BF, 0x0301, 0x03CC),  # ο + COMBINING ACUTE ACCENT = GREEK SMALL LETTER OMICRON WITH TONOS
+    (0x03C5, 0x0301, 0x03CD),  # υ + COMBINING ACUTE ACCENT = GREEK SMALL LETTER UPSILON WITH TONOS
+    (0x03C9, 0x0301, 0x03CE),  # ω + COMBINING ACUTE ACCENT = GREEK SMALL LETTER OMEGA WITH TONOS
+
+    # Common Cyrillic characters
+    (0x0418, 0x0306, 0x0419),  # И + COMBINING BREVE = CYRILLIC CAPITAL LETTER SHORT I
+    (0x0438, 0x0306, 0x0439),  # и + COMBINING BREVE = CYRILLIC SMALL LETTER SHORT I
+    (0x0423, 0x0306, 0x040E),  # У + COMBINING BREVE = CYRILLIC CAPITAL LETTER SHORT U
+    (0x0443, 0x0306, 0x045E),  # у + COMBINING BREVE = CYRILLIC SMALL LETTER SHORT U
+]
+
+full_recompose_description = "all possible recomposition pairs from the Unicode BMP"
+def collect_all_recomposition_pairs():
+    """Collect all possible recomposition pairs from the Unicode data."""
+    # Map to store recomposition pairs: (base, combining) -> recomposed
+    recompose_map = {}
+
+    # Process all assigned Unicode code points in BMP (Basic Multilingual Plane)
+    # We limit to BMP (0x0000-0xFFFF) to keep our table smaller with uint16_t
+    for cp in range(0, 0x10000):
+        try:
+            char = chr(cp)
+
+            # Skip unassigned or control characters
+            if not unicodedata.name(char, ''):
+                continue
+
+            # Find decomposition
+            decomp = unicodedata.decomposition(char)
+            if not decomp or '<' in decomp:  # Skip compatibility decompositions
+                continue
+
+            # Parse the decomposition
+            parts = decomp.split()
+            if len(parts) == 2:  # Simple base + combining mark
+                base = int(parts[0], 16)
+                combining = int(parts[1], 16)
+
+                # Only store if both are in BMP
+                if base < 0x10000 and combining < 0x10000:
+                    recompose_map[(base, combining)] = cp
+
+        except (ValueError, TypeError):
+            continue
+
+    # Convert to a list of tuples and sort for binary search
+    recompose_list = [(base, combining, recomposed)
+                     for (base, combining), recomposed in recompose_map.items()]
+    recompose_list.sort()
+
+    return recompose_list
+
+def validate_common_pairs(full_list):
+    """Validate that all common pairs are in the full list.
+
+    Raises:
+        ValueError: If any common pair is missing or has a different recomposition
+        value than what's in the full table.
+    """
+    full_pairs = {(base, combining): recomposed for base, combining, recomposed in full_list}
+    for base, combining, recomposed in COMMON_RECOMPOSITION_PAIRS:
+        full_recomposed = full_pairs.get((base, combining))
+        if full_recomposed is None:
+            error_msg = f"Error: Common pair (0x{base:04X}, 0x{combining:04X}) not found in full data"
+            print(error_msg)
+            raise ValueError(error_msg)
+        elif full_recomposed != recomposed:
+            error_msg = (f"Error: Common pair (0x{base:04X}, 0x{combining:04X}) has different recomposition: "
+                         f"0x{recomposed:04X} vs 0x{full_recomposed:04X}")
+            print(error_msg)
+            raise ValueError(error_msg)
+
+def generate_recomposition_table(use_full_list=False):
+    """Generate the recomposition table C code."""
+    # Output file name
+    c_file = "ucs_recompose.c"
+
+    # Get Unicode version information
+    unicode_version = unicodedata.unidata_version
+
+    # Collect all recomposition pairs for validation
+    full_recompose_list = collect_all_recomposition_pairs()
+
+    # Decide which list to use
+    if use_full_list:
+        print("Using full recomposition list...")
+        recompose_list = full_recompose_list
+        table_description = full_recompose_description
+        alt_list = COMMON_RECOMPOSITION_PAIRS
+        alt_description = common_recompose_description
+    else:
+        print("Using common recomposition list...")
+        # Validate that all common pairs are in the full list
+        validate_common_pairs(full_recompose_list)
+        recompose_list = sorted(COMMON_RECOMPOSITION_PAIRS)
+        table_description = common_recompose_description
+        alt_list = full_recompose_list
+        alt_description = full_recompose_description
+    generation_mode = " --full" if use_full_list else ""
+    alternative_mode = " --full" if not use_full_list else ""
+    table_description_detail = f"{table_description} ({len(recompose_list)} entries)"
+    alt_description_detail = f"{alt_description} ({len(alt_list)} entries)"
+
+    # Calculate min/max values for boundary checks
+    min_base = min(base for base, _, _ in recompose_list)
+    max_base = max(base for base, _, _ in recompose_list)
+    min_combining = min(combining for _, combining, _ in recompose_list)
+    max_combining = max(combining for _, combining, _ in recompose_list)
+
+    # Generate implementation file
+    with open(c_file, 'w') as f:
+        f.write(f"""\
+// SPDX-License-Identifier: GPL-2.0
+/*
+ * ucs_recompose.c - Unicode character recomposition
+ *
+ * Auto-generated by gen_ucs_recompose.py{generation_mode}
+ *
+ * Unicode Version: {unicode_version}
+ *
+{textwrap.fill(
+    f"This file contains a table with {table_description_detail}. " +
+    f"To generate a table with {alt_description_detail} instead, run:",
+    width=75, initial_indent=" * ", subsequent_indent=" * ")}
+ *
+ *   python gen_ucs_recompose.py{alternative_mode}
+ */
+
+#include <linux/types.h>
+#include <linux/array_size.h>
+#include <linux/bsearch.h>
+#include <linux/consolemap.h>
+
+/*
+ * Structure for recomposition pairs.
+ * First element is the base character, second is the combining mark,
+ * third is the recomposed character.
+ * Using uint16_t to save space since all values are within BMP range.
+ */
+struct recomposition {{
+	uint16_t base;
+	uint16_t combining;
+	uint16_t recomposed;
+}};
+
+/*
+ * Table of {table_description}
+ * Sorted by base character and then combining character for binary search
+ */
+static const struct recomposition recomposition_table[] = {{
+""")
+
+        # Write the recomposition table with comments
+        for base, combining, recomposed in recompose_list:
+            try:
+                base_name = unicodedata.name(chr(base))
+                combining_name = unicodedata.name(chr(combining))
+                recomposed_name = unicodedata.name(chr(recomposed))
+                comment = f"/* {base_name} + {combining_name} = {recomposed_name} */"
+            except ValueError:
+                comment = f"/* U+{base:04X} + U+{combining:04X} = U+{recomposed:04X} */"
+            f.write(f"\t{{ 0x{base:04X}, 0x{combining:04X}, 0x{recomposed:04X} }}, {comment}\n")
+
+        f.write(f"""\
+}};
+
+/*
+ * Boundary values for quick rejection
+ * These are calculated by analyzing the table during generation
+ */
+#define MIN_BASE_CHAR       0x{min_base:04X}
+#define MAX_BASE_CHAR       0x{max_base:04X}
+#define MIN_COMBINING_CHAR  0x{min_combining:04X}
+#define MAX_COMBINING_CHAR  0x{max_combining:04X}
+
+struct compare_key {{
+	uint16_t base;
+	uint16_t combining;
+}};
+
+static int recomposition_compare(const void *key, const void *element)
+{{
+	const struct compare_key *search_key = key;
+	const struct recomposition *table_entry = element;
+
+	/* Compare base character first */
+	if (search_key->base < table_entry->base)
+		return -1;
+	if (search_key->base > table_entry->base)
+		return 1;
+
+	/* Base characters match, now compare combining character */
+	if (search_key->combining < table_entry->combining)
+		return -1;
+	if (search_key->combining > table_entry->combining)
+		return 1;
+
+	/* Both match */
+	return 0;
+}}
+
+/**
+ * Attempt to recompose two Unicode characters into a single character.
+ *
+ * @param previous: Previous Unicode code point (UCS-4)
+ * @param current: Current Unicode code point (UCS-4)
+ * Return: Recomposed Unicode code point, or 0 if no recomposition is possible
+ */
+uint32_t ucs_recompose(uint32_t base, uint32_t combining)
+{{
+	/* Check if characters are within the range of our table */
+	if (base < MIN_BASE_CHAR || base > MAX_BASE_CHAR ||
+	    combining < MIN_COMBINING_CHAR || combining > MAX_COMBINING_CHAR)
+		return 0;
+
+	struct compare_key key = {{ base, combining }};
+
+	struct recomposition *result =
+		__inline_bsearch(&key, recomposition_table,
+				 ARRAY_SIZE(recomposition_table),
+				 sizeof(*recomposition_table),
+				 recomposition_compare);
+
+	return result ? result->recomposed : 0;
+}}
+""")
+
+if __name__ == "__main__":
+    parser = argparse.ArgumentParser(description="Generate Unicode recomposition table")
+    parser.add_argument("--full", action="store_true",
+                        help="Generate a full recomposition table (default: common pairs only)")
+    args = parser.parse_args()
+
+    generate_recomposition_table(use_full_list=args.full)