Make sorting stuff much faster

5 files changed, 64 insertions, 12 deletions

diff --git a/.gitignore b/.gitignore
index ed46773..e3baee2 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,5 @@
 fonts/
 done/
 out/
+bilingual_sort
 server
diff --git a/Makefile b/Makefile
new file mode 100644
index 0000000..e917176
--- /dev/null
+++ b/Makefile
@@ -0,0 +1,6 @@
+all:
+	go build server.go
+	${CC} -O3 bilingual_sort.c -o bilingual_sort
+
+clean:
+	rm -f bilingual_sort server
diff --git a/bilingual_sort.c b/bilingual_sort.c
new file mode 100644
index 0000000..0ea5e0c
--- /dev/null
+++ b/bilingual_sort.c
@@ -0,0 +1,41 @@
+/* This is a simple program that runs in an infinite loop.  It reads strings
+ * from standard input in the form of “str1:str2”, sorts them according to the
+ * locale (set by gawk), and then prints out the first string in sorted order.
+ *
+ * The reason we use this instead of just sort(1) or something is because
+ * spawning one process for each comparison is super slow.  This lets us do
+ * multiple comparisons of string-pairs in only a single process.
+ */
+
+#define _GNU_SOURCE
+
+#include <locale.h>
+#include <stdbool.h>
+#include <stdio.h>
+#include <stdlib.h>
+#include <string.h>
+
+int
+main(void)
+{
+	char *s1, *s2,
+	     line[256];
+
+	/* We need line buffering to make gawk interact with this properly */
+	setvbuf(stdout, NULL, _IOLBF, 0);
+	setlocale(LC_ALL, "");
+
+	while (true) {
+		if (fgets(line, sizeof(line), stdin) == NULL)
+			break;
+
+		s1 = line;
+		s2 = strchr(line, ':') + 1;
+		s2[-1] = '\0';
+		*strchrnul(s2, '\n') = '\0';
+
+		puts(strcoll(s1, s2) < 0 ? s1 : s2);
+	}
+
+	return EXIT_SUCCESS;
+}
diff --git a/build b/build
index 66cf511..1a4d07e 100755
--- a/build
+++ b/build
@@ -221,10 +221,13 @@ compile_for_lang() {
 			}
 
 			END {
-				if (lang == "en")
-					asort(col)
-				else
-					asort(col, col, "bilingual_sort")
+				locale = lang == "pt" \
+					? "pt_BR.UTF-8"   \
+					: lang "_" toupper(lang) ".UTF-8"
+
+				CMD = "LC_ALL=" locale " ./bilingual_sort"
+				asort(col, col, "bilingual_sort")
+				close(CMD)
 
 				for (i = 1; i <= 3; i++) {
 					if (i == 3 && denom == 5) {
@@ -289,10 +292,13 @@ compile_for_lang() {
 			}
 
 			END {
-				if (lang == "en")
-					asort(col)
-				else
-					asort(col, col, "bilingual_sort")
+				locale = lang == "pt" \
+					? "pt_BR.UTF-8"   \
+					: lang "_" toupper(lang) ".UTF-8"
+
+				CMD = "LC_ALL=" locale " ./bilingual_sort"
+				asort(col, col, "bilingual_sort")
+				close(CMD)
 
 				for (i = 1; i <= 2; i++) {
 					if (i == 1)
diff --git a/scripts/bilingual_sort.awk b/scripts/bilingual_sort.awk
index 6f6f53a..7e50136 100644
--- a/scripts/bilingual_sort.awk
+++ b/scripts/bilingual_sort.awk
@@ -3,10 +3,8 @@ function bilingual_sort(i1, v1, i2, v2,    l, r, f)
 	l = cc2name[v1[0]]
 	r = cc2name[v2[0]]
 
-	printf "%s\n%s\n", l, r |& "sort"
-	close("sort", "to")
-	"sort" |& getline f
-	close("sort", "from")
+	print l ":" r |& CMD
+	CMD |& getline f
 
 	return f == l ? -1 : +1
 }