Instead of 66 objects, have 1 object. More confusing, but faster?

This commit is contained in:
Neil 2022-12-15 23:18:11 -08:00
parent 2807b5704d
commit 5d3e027f36
1 changed files with 67 additions and 61 deletions

View File

@ -19,33 +19,37 @@
#include "../src/array.h" #include "../src/array.h"
/** Append a text file, `fn`, to `c`, and add a '\0'. /** Append a text file, `fn`, to `c`, and add a '\0'.
@return Success. A partial read is failure. @throws[fopen, fread, malloc] @return The start of the appended file or null on error. A partial read is a
failure. @throws[fopen, fread, malloc]
@throws[EISEQ] The text file has embedded nulls. @throws[EISEQ] The text file has embedded nulls.
@throws[ERANGE] If the standard library does not follow POSIX. */ @throws[ERANGE] If the standard library does not follow POSIX. */
static int append_file(struct char_array *c, const char *const fn) { static char *append_file(struct char_array *text, const char *const fn) {
FILE *fp = 0; FILE *fp = 0;
const size_t granularity = 1024; const size_t granularity = 1024;
size_t nread; size_t nread, start;
char *cursor; char *cursor;
int success = 0; int success = 1;
assert(c && fn); assert(text && fn);
start = text->size;
if(!(fp = fopen(fn, "r"))) goto catch; if(!(fp = fopen(fn, "r"))) goto catch;
/* Read entire file in chunks. */ /* Read entire file in chunks. */
do if(!(cursor = char_array_buffer(c, granularity)) do if(!(cursor = char_array_buffer(text, granularity))
|| (nread = fread(cursor, 1, granularity, fp), ferror(fp)) || (nread = fread(cursor, 1, granularity, fp), ferror(fp))
|| !char_array_append(c, nread)) goto catch; || !char_array_append(text, nread)) goto catch;
while(nread == granularity); while(nread == granularity);
/* File to `C` string. */ /* File to `C` string. */
if(!(cursor = char_array_new(c))) goto catch; if(!(cursor = char_array_new(text))) goto catch;
*cursor = '\0'; *cursor = '\0';
/* Binary files with embedded '\0' are not allowed. */ /* Binary files with embedded '\0' are not allowed; check just this read. */
if(strchr(c->data, '\0') != cursor) { errno = EILSEQ; goto catch; } if(strchr(text->data + start, '\0') != cursor)
{ success = 1; goto finally; } { errno = EILSEQ; goto catch; }
goto finally;
catch: catch:
if(!errno) errno = EILSEQ; /* Will never be true on POSIX. */ if(!errno) errno = EILSEQ; /* Will never be true on POSIX. */
success = 0;
finally: finally:
if(fp) fclose(fp); if(fp) fclose(fp);
return success; return success ? text->data + start : 0;
} }
@ -245,12 +249,12 @@ static uint32_t lowbias32_r(uint32_t x) {
return x; return x;
} }
/** Two hash-tables use the same structure. */
union kjvcite { union kjvcite {
/* Overkill, but no initializing unused bits, 12 + 13 + 7 = 32. */ /* Overkill, but no initializing unused bits, 12 + 13 + 7 = 32. */
struct { unsigned verse : 12, chapter : 13, book : 7; }; struct { unsigned verse : 12, chapter : 13, book : 7; };
uint32_t u32; uint32_t u32;
}; };
static uint32_t kjv_hash(const union kjvcite x) { return lowbias32(x.u32); } static uint32_t kjv_hash(const union kjvcite x) { return lowbias32(x.u32); }
static union kjvcite kjv_unhash(const uint32_t x) { static union kjvcite kjv_unhash(const uint32_t x) {
union kjvcite k; union kjvcite k;
@ -261,25 +265,25 @@ static void kjv_to_string(const union kjvcite x, char (*const a)[12])
{ sprintf(*a, "%.4s%u:%u", kjv_book_string[x.book], { sprintf(*a, "%.4s%u:%u", kjv_book_string[x.book],
(x.chapter + 1) % 1000, (x.verse + 1) % 1000); } (x.chapter + 1) % 1000, (x.verse + 1) % 1000); }
static uint32_t kjvword_hash(const union kjvcite x) { return kjv_hash(x); } /** Derived information on verse word count. */
static union kjvcite kjvword_unhash(const uint32_t x) { return kjv_unhash(x); } static uint32_t verse_hash(const union kjvcite x) { return kjv_hash(x); }
static void kjvword_to_string(const union kjvcite x, char (*const a)[12]) static union kjvcite verse_unhash(const uint32_t x) { return kjv_unhash(x); }
static void verse_to_string(const union kjvcite x, char (*const a)[12])
{ kjv_to_string(x, a); } { kjv_to_string(x, a); }
#define TABLE_NAME verse
#define TABLE_NAME kjvword
#define TABLE_KEY union kjvcite #define TABLE_KEY union kjvcite
#define TABLE_UINT uint32_t #define TABLE_UINT uint32_t
#define TABLE_VALUE unsigned #define TABLE_VALUE unsigned
#define TABLE_DEFAULT 0
#define TABLE_INVERSE #define TABLE_INVERSE
#define TABLE_DEFAULT 0
#define TABLE_TO_STRING #define TABLE_TO_STRING
#include "../src/table.h" #include "../src/table.h"
/* A set of verses. */
static uint32_t kjvset_hash(const union kjvcite x) { return kjv_hash(x); } static uint32_t kjvset_hash(const union kjvcite x) { return kjv_hash(x); }
static union kjvcite kjvset_unhash(const uint32_t x) { return kjv_unhash(x); } static union kjvcite kjvset_unhash(const uint32_t x) { return kjv_unhash(x); }
static void kjvset_to_string(const union kjvcite x, char (*const a)[12]) static void kjvset_to_string(const union kjvcite x, char (*const a)[12])
{ kjv_to_string(x, a); } { kjv_to_string(x, a); }
#define TABLE_NAME kjvset #define TABLE_NAME kjvset
#define TABLE_KEY union kjvcite #define TABLE_KEY union kjvcite
#define TABLE_UINT uint32_t #define TABLE_UINT uint32_t
@ -288,76 +292,78 @@ static void kjvset_to_string(const union kjvcite x, char (*const a)[12])
#include "../src/table.h" #include "../src/table.h"
int main(void) { int main(void) {
const char *const dir_name = "KJV"; const char *const dir_kjv = "KJV";
struct char_array kjv[KJV_BOOK_SIZE] = { 0 }; struct {
struct kjvword_table words = { 0 }; struct char_array backing;
int success = EXIT_SUCCESS; struct verse_table verses;
size_t words;
} kjv = { 0 };
DIR *dir = 0; DIR *dir = 0;
struct dirent *de = 0; struct dirent *de = 0;
unsigned i; struct { size_t offset; int is; } build[KJV_BOOK_SIZE] = { 0 };
size_t cum_words = 0; enum kjv_book b = 0;
int success = EXIT_SUCCESS, attempted_closedir = 0;
errno = 0; errno = 0;
/* Read in the kjv from all files. This is overkill, we don't need to keep /* For all files in directory KJV with <#>*.txt, read into backing. */
all the data, just count. Maybe we'll do something else later? */ if(chdir(dir_kjv) == -1 || !(dir = opendir("."))) goto catch;
if(chdir(dir_name) == -1 || !(dir = opendir("."))) goto catch; while((de = readdir(dir))) {
while((de = readdir(dir))) { /* For all files in directory. */
unsigned ordinal; unsigned ordinal;
enum kjv_book b; char *unstable_book;
if(!kjv_filename(de->d_name, &ordinal)) /* Extract ordinal. */ if(!kjv_filename(de->d_name, &ordinal)) continue; /* Extract no. */
{ /*fprintf(stderr, "Ignored <%s>.\n", de->d_name);*/ continue; }
/*fprintf(stderr, "<%s> ordinal: %u\n", de->d_name, ordinal);*/ /*fprintf(stderr, "<%s> ordinal: %u\n", de->d_name, ordinal);*/
if(ordinal < 1 || ordinal > KJV_BOOK_SIZE) if(ordinal < 1 || ordinal > KJV_BOOK_SIZE)
{ errno = ERANGE; goto catch; } /* Not in range. */ { errno = ERANGE; goto catch; } /* Not in range. */
if(kjv[b = ordinal - 1].data) /* Convert to zero-based. */ if(build[b = ordinal - 1].is) /* Convert to zero-based. */
{ errno = EDOM; goto catch; } /* Duplicate. */ { errno = EDOM; goto catch; } /* Duplicate. */
if(!append_file(kjv + b, de->d_name)) goto catch; if(!(unstable_book = append_file(&kjv.backing, de->d_name))) goto catch;
build[b].is = 1;
build[b].offset = (size_t)(unstable_book - kjv.backing.data);
} }
closedir(dir), de = 0, dir = 0; if(attempted_closedir = 1, closedir(dir) == -1) goto catch; dir = 0;
/* Parse number of words in each verse. */ /* Now backing is stable; count all the words for each verse. */
for(i = 0; i < KJV_BOOK_SIZE; i++) { for(b = 0; b < KJV_BOOK_SIZE; b++) {
struct lex x = lex(kjv[i].data); struct lex x;
if(!x.cursor) { fprintf(stderr, "Missing book [%u]%s.\n", if(!build[b].is) { fprintf(stderr, "Missing book [%u]%s.\n",
i + 1, kjv_book_string[i]); errno = EDOM; goto catch; } b + 1, kjv_book_string[b]); errno = EDOM; goto catch; }
/*printf("[%u]%s: cumulative %zu.\n", x = lex(kjv.backing.data + build[b].offset);
i + 1, kjv_book_string[i], cum_words);*/
while(lex_next_verse(&x)) { while(lex_next_verse(&x)) {
const union kjvcite c const union kjvcite cite
= { .verse = x.verse, .chapter = x.chapter, .book = i }; = { .book = b, .chapter = x.chapter, .verse = x.verse };
unsigned *w; unsigned *words;
switch(kjvword_table_assign(&words, c, &w)) { switch(verse_table_assign(&kjv.verses, cite, &words)) {
case TABLE_PRESENT: fprintf(stderr, "[%u]%s %u:%u duplicated.\n", case TABLE_PRESENT: fprintf(stderr, "[%u]%s %u:%u duplicated.\n",
i + 1, kjv_book_string[i], x.chapter, x.verse); errno = EDOM; b + 1, kjv_book_string[b], x.chapter, x.verse); errno = EDOM;
case TABLE_ERROR: goto catch; case TABLE_ERROR: goto catch;
case TABLE_ABSENT: *w = x.words; break; case TABLE_ABSENT: break;
} }
cum_words += x.words; *words = x.words, kjv.words += x.words;
} }
if(x.error) { fprintf(stderr, "[%u]%s on line %zu\n", if(x.error) { fprintf(stderr, "[%u]%s on line %zu\n",
i + 1, kjv_book_string[i], x.line); goto catch; } b + 1, kjv_book_string[b], x.line); goto catch; }
} }
printf("words: %s\n", kjvword_table_to_string(&words)); printf("words: %s\n", verse_table_to_string(&kjv.verses));
printf("kjv: %zu total words\n", cum_words); printf("kjv: %zu total words\n", kjv.words);
{ {
union kjvcite c; union kjvcite c;
struct kjvword_table_iterator it = kjvword_table_begin(&words); struct verse_table_iterator it = verse_table_begin(&kjv.verses);
unsigned *w; unsigned *w;
while(kjvword_table_next(&it, &c, &w)) while(verse_table_next(&it, &c, &w))
printf("%s %u:%u -> %u\n", printf("%s %u:%u -> %u\n",
kjv_book_string[c.book], c.chapter, c.verse, *w); kjv_book_string[c.book], c.chapter, c.verse, *w);
c = (union kjvcite){ .book = Genesis, .chapter = 1, .verse = 1 }; c = (union kjvcite){ .book = Genesis, .chapter = 1, .verse = 1 };
printf("1:1:1 -> %u\n", kjvword_table_get(&words, c)); printf("1:1:1 -> %u\n", verse_table_get(&kjv.verses, c));
} }
goto finally; goto finally;
catch: catch:
success = EXIT_FAILURE; success = EXIT_FAILURE;
perror(de ? de->d_name : dir_name); if(de) fprintf(stderr, "While reading %s.\n", de->d_name);
if(dir && closedir(dir)) perror(dir_name); perror(de ? de->d_name : dir_kjv);
if(dir && !attempted_closedir && closedir(dir) == -1) perror(dir_kjv);
finally: finally:
/*kjvset_table_();*/ verse_table_(&kjv.verses);
kjvword_table_(&words); char_array_(&kjv.backing);
for(i = 0; i < KJV_BOOK_SIZE; i++) char_array_(kjv + i);
return success; return success;
} }