1 3b0f3d61 2020-01-22 neels /* Split source by line breaks, and calculate a simplistic checksum. */
  2 3b0f3d61 2020-01-22 neels /*
  3 3b0f3d61 2020-01-22 neels  * Copyright (c) 2020 Neels Hofmeyr <neels@hofmeyr.de>
  4 3b0f3d61 2020-01-22 neels  *
  5 3b0f3d61 2020-01-22 neels  * Permission to use, copy, modify, and distribute this software for any
  6 3b0f3d61 2020-01-22 neels  * purpose with or without fee is hereby granted, provided that the above
  7 3b0f3d61 2020-01-22 neels  * copyright notice and this permission notice appear in all copies.
  8 3b0f3d61 2020-01-22 neels  *
  9 3b0f3d61 2020-01-22 neels  * THE SOFTWARE IS PROVIDED "AS IS" AND THE AUTHOR DISCLAIMS ALL WARRANTIES
 10 3b0f3d61 2020-01-22 neels  * WITH REGARD TO THIS SOFTWARE INCLUDING ALL IMPLIED WARRANTIES OF
 11 3b0f3d61 2020-01-22 neels  * MERCHANTABILITY AND FITNESS. IN NO EVENT SHALL THE AUTHOR BE LIABLE FOR
 12 3b0f3d61 2020-01-22 neels  * ANY SPECIAL, DIRECT, INDIRECT, OR CONSEQUENTIAL DAMAGES OR ANY DAMAGES
 13 3b0f3d61 2020-01-22 neels  * WHATSOEVER RESULTING FROM LOSS OF USE, DATA OR PROFITS, WHETHER IN AN
 14 3b0f3d61 2020-01-22 neels  * ACTION OF CONTRACT, NEGLIGENCE OR OTHER TORTIOUS ACTION, ARISING OUT OF
 15 3b0f3d61 2020-01-22 neels  * OR IN CONNECTION WITH THE USE OR PERFORMANCE OF THIS SOFTWARE.
 16 3b0f3d61 2020-01-22 neels  */
 17 3b0f3d61 2020-01-22 neels 
 18 e10a628a 2020-09-16 stsp #include <errno.h>
 19 e10a628a 2020-09-16 stsp #include <stdbool.h>
 20 fe6d58fb 2020-11-14 naddy #include <stdint.h>
 21 c6eecea3 2020-07-26 stsp #include <stdio.h>
 22 e10a628a 2020-09-16 stsp #include <stdlib.h>
 23 e10a628a 2020-09-16 stsp #include <unistd.h>
 24 845f3575 2020-10-22 neels #include <ctype.h>
 25 c6eecea3 2020-07-26 stsp 
 26 1dfba055 2020-10-07 stsp #include <arraylist.h>
 27 1dfba055 2020-10-07 stsp #include <diff_main.h>
 28 1dfba055 2020-10-07 stsp 
 29 85ab4559 2020-09-22 stsp #include "diff_internal.h"
 30 2a1b94d0 2020-09-26 stsp #include "diff_debug.h"
 31 3b0f3d61 2020-01-22 neels 
 32 f8b2e31e 2022-08-03 stsp unsigned int
 33 29916bb6 2020-11-18 stsp diff_atom_hash_update(unsigned int hash, unsigned char atom_byte)
 34 29916bb6 2020-11-18 stsp {
 35 29916bb6 2020-11-18 stsp 	return hash * 23 + atom_byte;
 36 29916bb6 2020-11-18 stsp }
 37 29916bb6 2020-11-18 stsp 
 38 61a7b578 2020-05-06 neels static int
 39 c6eecea3 2020-07-26 stsp diff_data_atomize_text_lines_fd(struct diff_data *d)
 40 3b0f3d61 2020-01-22 neels {
 41 7a54ad3a 2020-09-20 stsp 	off_t pos = 0;
 42 c6eecea3 2020-07-26 stsp 	const off_t end = pos + d->len;
 43 c6eecea3 2020-07-26 stsp 	unsigned int array_size_estimate = d->len / 50;
 44 c6eecea3 2020-07-26 stsp 	unsigned int pow2 = 1;
 45 845f3575 2020-10-22 neels 	bool ignore_whitespace = (d->diff_flags & DIFF_FLAG_IGNORE_WHITESPACE);
 46 e51ebd83 2020-11-21 stsp 	bool embedded_nul = false;
 47 7a54ad3a 2020-09-20 stsp 
 48 c6eecea3 2020-07-26 stsp 	while (array_size_estimate >>= 1)
 49 c6eecea3 2020-07-26 stsp 		pow2++;
 50 c6eecea3 2020-07-26 stsp 
 51 c6eecea3 2020-07-26 stsp 	ARRAYLIST_INIT(d->atoms, 1 << pow2);
 52 c6eecea3 2020-07-26 stsp 
 53 7a54ad3a 2020-09-20 stsp 	if (fseek(d->root->f, 0L, SEEK_SET) == -1)
 54 7a54ad3a 2020-09-20 stsp 		return errno;
 55 7a54ad3a 2020-09-20 stsp 
 56 c6eecea3 2020-07-26 stsp 	while (pos < end) {
 57 c6eecea3 2020-07-26 stsp 		off_t line_end = pos;
 58 c6eecea3 2020-07-26 stsp 		unsigned int hash = 0;
 59 c6eecea3 2020-07-26 stsp 		unsigned char buf[512];
 60 7a54ad3a 2020-09-20 stsp 		size_t r, i;
 61 c6eecea3 2020-07-26 stsp 		struct diff_atom *atom;
 62 c6eecea3 2020-07-26 stsp 		int eol = 0;
 63 c6eecea3 2020-07-26 stsp 
 64 c6eecea3 2020-07-26 stsp 		while (eol == 0 && line_end < end) {
 65 7a54ad3a 2020-09-20 stsp 			r = fread(buf, sizeof(char), sizeof(buf), d->root->f);
 66 7a54ad3a 2020-09-20 stsp 			if (r == 0 && ferror(d->root->f))
 67 674563ab 2022-10-11 tj 				return EIO;
 68 c6eecea3 2020-07-26 stsp 			i = 0;
 69 c6eecea3 2020-07-26 stsp 			while (eol == 0 && i < r) {
 70 c6eecea3 2020-07-26 stsp 				if (buf[i] != '\r' && buf[i] != '\n') {
 71 845f3575 2020-10-22 neels 					if (!ignore_whitespace
 72 1dce05e8 2022-11-17 op 					    || !isspace((unsigned char)buf[i]))
 73 29916bb6 2020-11-18 stsp 						hash = diff_atom_hash_update(
 74 29916bb6 2020-11-18 stsp 						    hash, buf[i]);
 75 e51ebd83 2020-11-21 stsp 					if (buf[i] == '\0')
 76 e51ebd83 2020-11-21 stsp 						embedded_nul = true;
 77 c6eecea3 2020-07-26 stsp 					line_end++;
 78 c6eecea3 2020-07-26 stsp 				} else
 79 c6eecea3 2020-07-26 stsp 					eol = buf[i];
 80 c6eecea3 2020-07-26 stsp 				i++;
 81 c6eecea3 2020-07-26 stsp 			}
 82 c6eecea3 2020-07-26 stsp 		}
 83 c6eecea3 2020-07-26 stsp 
 84 c6eecea3 2020-07-26 stsp 		/* When not at the end of data, the line ending char ('\r' or
 85 c6eecea3 2020-07-26 stsp 		 * '\n') must follow */
 86 c6eecea3 2020-07-26 stsp 		if (line_end < end)
 87 c6eecea3 2020-07-26 stsp 			line_end++;
 88 c6eecea3 2020-07-26 stsp 		/* If that was an '\r', also pull in any following '\n' */
 89 c6eecea3 2020-07-26 stsp 		if (line_end < end && eol == '\r') {
 90 7a54ad3a 2020-09-20 stsp 			if (fseeko(d->root->f, line_end, SEEK_SET) == -1)
 91 c6eecea3 2020-07-26 stsp 				return errno;
 92 7a54ad3a 2020-09-20 stsp 			r = fread(buf, sizeof(char), sizeof(buf), d->root->f);
 93 7a54ad3a 2020-09-20 stsp 			if (r == 0 && ferror(d->root->f))
 94 674563ab 2022-10-11 tj 				return EIO;
 95 ed9312f0 2022-07-26 op 			if (r > 0 && buf[0] == '\n')
 96 c6eecea3 2020-07-26 stsp 				line_end++;
 97 c6eecea3 2020-07-26 stsp 		}
 98 c6eecea3 2020-07-26 stsp 
 99 c6eecea3 2020-07-26 stsp 		/* Record the found line as diff atom */
100 c6eecea3 2020-07-26 stsp 		ARRAYLIST_ADD(atom, d->atoms);
101 c6eecea3 2020-07-26 stsp 		if (!atom)
102 3e6cba3a 2020-08-13 stsp 			return ENOMEM;
103 c6eecea3 2020-07-26 stsp 
104 c6eecea3 2020-07-26 stsp 		*atom = (struct diff_atom){
105 ad5b3f85 2020-10-12 neels 			.root = d,
106 c6eecea3 2020-07-26 stsp 			.pos = pos,
107 c6eecea3 2020-07-26 stsp 			.at = NULL,	/* atom data is not memory-mapped */
108 c6eecea3 2020-07-26 stsp 			.len = line_end - pos,
109 c6eecea3 2020-07-26 stsp 			.hash = hash,
110 c6eecea3 2020-07-26 stsp 		};
111 c6eecea3 2020-07-26 stsp 
112 c6eecea3 2020-07-26 stsp 		/* Starting point for next line: */
113 c6eecea3 2020-07-26 stsp 		pos = line_end;
114 7a54ad3a 2020-09-20 stsp 		if (fseeko(d->root->f, pos, SEEK_SET) == -1)
115 03f49727 2020-09-20 stsp 			return errno;
116 c6eecea3 2020-07-26 stsp 	}
117 c6eecea3 2020-07-26 stsp 
118 e51ebd83 2020-11-21 stsp 	/* File are considered binary if they contain embedded '\0' bytes. */
119 e51ebd83 2020-11-21 stsp 	if (embedded_nul)
120 e51ebd83 2020-11-21 stsp 		d->atomizer_flags |= DIFF_ATOMIZER_FOUND_BINARY_DATA;
121 e51ebd83 2020-11-21 stsp 
122 c6eecea3 2020-07-26 stsp 	return DIFF_RC_OK;
123 c6eecea3 2020-07-26 stsp }
124 c6eecea3 2020-07-26 stsp 
125 c6eecea3 2020-07-26 stsp static int
126 c6eecea3 2020-07-26 stsp diff_data_atomize_text_lines_mmap(struct diff_data *d)
127 c6eecea3 2020-07-26 stsp {
128 3b0f3d61 2020-01-22 neels 	const uint8_t *pos = d->data;
129 3b0f3d61 2020-01-22 neels 	const uint8_t *end = pos + d->len;
130 845f3575 2020-10-22 neels 	bool ignore_whitespace = (d->diff_flags & DIFF_FLAG_IGNORE_WHITESPACE);
131 e51ebd83 2020-11-21 stsp 	bool embedded_nul = false;
132 3b0f3d61 2020-01-22 neels 	unsigned int array_size_estimate = d->len / 50;
133 3b0f3d61 2020-01-22 neels 	unsigned int pow2 = 1;
134 3b0f3d61 2020-01-22 neels 	while (array_size_estimate >>= 1)
135 3b0f3d61 2020-01-22 neels 		pow2++;
136 3b0f3d61 2020-01-22 neels 
137 3b0f3d61 2020-01-22 neels 	ARRAYLIST_INIT(d->atoms, 1 << pow2);
138 3b0f3d61 2020-01-22 neels 
139 3b0f3d61 2020-01-22 neels 	while (pos < end) {
140 3b0f3d61 2020-01-22 neels 		const uint8_t *line_end = pos;
141 3b0f3d61 2020-01-22 neels 		unsigned int hash = 0;
142 3b0f3d61 2020-01-22 neels 
143 3b0f3d61 2020-01-22 neels 		while (line_end < end && *line_end != '\r' && *line_end != '\n') {
144 845f3575 2020-10-22 neels 			if (!ignore_whitespace
145 1dce05e8 2022-11-17 op 			    || !isspace((unsigned char)*line_end))
146 cd9ef01a 2022-08-03 op 				hash = diff_atom_hash_update(hash, *line_end);
147 e51ebd83 2020-11-21 stsp 			if (*line_end == '\0')
148 e51ebd83 2020-11-21 stsp 				embedded_nul = true;
149 3b0f3d61 2020-01-22 neels 			line_end++;
150 3b0f3d61 2020-01-22 neels 		}
151 3b0f3d61 2020-01-22 neels 
152 0d27172a 2020-05-06 neels 		/* When not at the end of data, the line ending char ('\r' or
153 0d27172a 2020-05-06 neels 		 * '\n') must follow */
154 ed9312f0 2022-07-26 op 		if (line_end < end && *line_end == '\r')
155 3b0f3d61 2020-01-22 neels 			line_end++;
156 ed9312f0 2022-07-26 op 		if (line_end < end && *line_end == '\n')
157 3b0f3d61 2020-01-22 neels 			line_end++;
158 3b0f3d61 2020-01-22 neels 
159 3b0f3d61 2020-01-22 neels 		/* Record the found line as diff atom */
160 3b0f3d61 2020-01-22 neels 		struct diff_atom *atom;
161 3b0f3d61 2020-01-22 neels 		ARRAYLIST_ADD(atom, d->atoms);
162 3b0f3d61 2020-01-22 neels 		if (!atom)
163 3e6cba3a 2020-08-13 stsp 			return ENOMEM;
164 3b0f3d61 2020-01-22 neels 
165 3b0f3d61 2020-01-22 neels 		*atom = (struct diff_atom){
166 ad5b3f85 2020-10-12 neels 			.root = d,
167 c6eecea3 2020-07-26 stsp 			.pos = (off_t)(pos - d->data),
168 3b0f3d61 2020-01-22 neels 			.at = pos,
169 3b0f3d61 2020-01-22 neels 			.len = line_end - pos,
170 3b0f3d61 2020-01-22 neels 			.hash = hash,
171 3b0f3d61 2020-01-22 neels 		};
172 3b0f3d61 2020-01-22 neels 
173 3b0f3d61 2020-01-22 neels 		/* Starting point for next line: */
174 3b0f3d61 2020-01-22 neels 		pos = line_end;
175 3b0f3d61 2020-01-22 neels 	}
176 3b0f3d61 2020-01-22 neels 
177 e51ebd83 2020-11-21 stsp 	/* File are considered binary if they contain embedded '\0' bytes. */
178 e51ebd83 2020-11-21 stsp 	if (embedded_nul)
179 e51ebd83 2020-11-21 stsp 		d->atomizer_flags |= DIFF_ATOMIZER_FOUND_BINARY_DATA;
180 e51ebd83 2020-11-21 stsp 
181 3b0f3d61 2020-01-22 neels 	return DIFF_RC_OK;
182 3b0f3d61 2020-01-22 neels }
183 3b0f3d61 2020-01-22 neels 
184 c6eecea3 2020-07-26 stsp static int
185 c6eecea3 2020-07-26 stsp diff_data_atomize_text_lines(struct diff_data *d)
186 c6eecea3 2020-07-26 stsp {
187 40dba3d8 2020-10-16 stsp 	if (d->data == NULL)
188 1ea18522 2020-10-16 stsp 		return diff_data_atomize_text_lines_fd(d);
189 40dba3d8 2020-10-16 stsp 	else
190 40dba3d8 2020-10-16 stsp 		return diff_data_atomize_text_lines_mmap(d);
191 c6eecea3 2020-07-26 stsp }
192 c6eecea3 2020-07-26 stsp 
193 3e6cba3a 2020-08-13 stsp int
194 c16dde50 2020-10-22 stsp diff_atomize_text_by_line(void *func_data, struct diff_data *d)
195 3b0f3d61 2020-01-22 neels {
196 c16dde50 2020-10-22 stsp 	return diff_data_atomize_text_lines(d);
197 3b0f3d61 2020-01-22 neels }