sfeed

commit ace0f818d172c81e23783b9c7e571464dcd0f604
parent 6446070da557bf8b56fa44b2bbdc4690edf490a9
Author: Hiltjo Posthuma <hiltjo@codemadness.org>
Date:   Wed, 22 Aug 2018 16:29:20 +0200

sfeed_tail: improvements

- sfeed_tail only reads from files now, not from stdin anymore. This had too many
  caveats.
- Instead of a timer it now detects changes faster and is more efficient using
  memory.
- Improve documentation of its behaviour.

Diffstat:
M
sfeed_tail.1
 | 
35
++++++++++++++---------------------
M
sfeed_tail.c
 | 
62
+++++++++++++++++++++++++++++++-------------------------------

2 files changed, 45 insertions(+), 52 deletions(-)
diff --git a/sfeed_tail.1 b/sfeed_tail.1
@@ -1,4 +1,4 @@
-.Dd January 29, 2016
+.Dd August 22, 2018
 .Dt SFEED_TAIL 1
 .Os
 .Sh NAME
@@ -6,40 +6,33 @@
 .Nd format unseen feed data to a plain-text list
 .Sh SYNOPSIS
 .Nm
-.Op Ar file...
+.Ar file...
 .Sh DESCRIPTION
 .Nm
-formats unseen feed data (TSV) from
-.Xr sfeed 1
-from stdin or
+formats only new and unseen feed data (TSV) from
+one or more
 .Ar file
 to stdout as a plain-text list.
-If one or more
-.Ar file
-are specified, the basename of the
+The basename of the
 .Ar file
 is used as the feed name in the output.
-If no
-.Ar file
-parameters are specified and so the data is read from stdin the feed name
-is empty.
-.Pp
-.Nm
-will mark the initial items on the first run as seen.
-Then it will print the initial items if data is read from stdin, if the data
-is read on the first run by specifying the argument(s)
-.Ar file
-it will not show the initial seen items.
 .Pp
 Unseen items are printed per line in a similar format to
 .Xr sfeed_plain 1 ,
 duplicate items are ignored.
-The list of unique items is determined by the fields: feedname, item id and
+The list of unique items is determined by the fields: item id, item link and
 UNIX timestamp of the item date.
+.Pp
+.Nm
+will also only process and show items that are considered new: the item
+timestamp is not older than a day ago.
 .Sh IMPLEMENTATION NOTES
+.Nm
+checks for file modifications each second by checking the filesize and
+modification time.
 Keep in mind that because
 .Nm
-keeps a list of all the items it can consume much memory.
+keeps a list of items it can potentially consume much memory.
 .Sh SEE ALSO
 .Xr sfeed 1 ,
 .Xr sfeed_plain 1 ,
diff --git a/sfeed_tail.c b/sfeed_tail.c
@@ -1,3 +1,4 @@
+#include <sys/stat.h>
 #include <sys/types.h>
 
 #include <ctype.h>
@@ -11,11 +12,10 @@
 #include "tree.h"
 #include "util.h"
 
-static int firsttime;
-static int sleepsecs;
 static char *line;
 static size_t linesize;
-time_t comparetime;
+static int changed;
+static time_t comparetime;
 
 struct line {
 	char *id;
@@ -47,8 +47,6 @@ gc(void)
 
 	RB_FOREACH_SAFE(line, linetree, &head, tmp) {
 		if (line->timestamp < comparetime) {
-/*			printf("DEBUG: gc: removing: %s %s\n",
-				line->id, line->title);*/
 			free(line->id);
 			free(line->link);
 			free(line->title);
@@ -90,8 +88,7 @@ printfeed(FILE *fp, const char *feedname)
 		if (RB_FIND(linetree, &head, &search))
 			continue;
 
-/*		printf("DEBUG: new: id: %s, link: %s, title: %s\n",
-			fields[FieldId], fields[FieldLink], fields[FieldTitle]);*/
+		changed = 1;
 
 		if (!(add = calloc(1, sizeof(*add))))
 			err(1, "calloc");
@@ -104,9 +101,6 @@ printfeed(FILE *fp, const char *feedname)
 		add->timestamp = parsedtime;
 		RB_INSERT(linetree, &head, add);
 
-		if (firsttime)
-			continue;
-
 		if (feedname[0]) {
 			printutf8pad(stdout, feedname, 15, ' ');
 			fputs("  ", stdout);
@@ -123,6 +117,7 @@ printfeed(FILE *fp, const char *feedname)
 int
 main(int argc, char *argv[])
 {
+	struct stat *stfiles, st;
 	char *name;
 	FILE *fp;
 	int i, slept = 0;
@@ -130,45 +125,50 @@ main(int argc, char *argv[])
 	if (pledge("stdio rpath", NULL) == -1)
 		err(1, "pledge");
 
+	if (argc <= 1) {
+		fprintf(stderr, "usage: %s <file>...\n", argv[0]);
+		return 1;
+	}
+
 	setlocale(LC_CTYPE, "");
 
-	if (pledge(argc == 1 ? "stdio" : "stdio rpath", NULL) == -1)
-		err(1, "pledge");
+	if (!(stfiles = calloc(argc - 1, sizeof(*stfiles))))
+		err(1, "calloc");
 
-	if (argc == 1)
-		sleepsecs = 1;
-	else
-		sleepsecs = 300;
+	while (1) {
+		changed = 0;
 
-	for (firsttime = (argc > 1); ; firsttime = 0) {
 		if ((comparetime = time(NULL)) == -1)
 			err(1, "time");
 		/* 1 day is old news */
 		comparetime -= 86400;
-		if (argc == 1) {
-			printfeed(stdin, "");
-		} else {
-			for (i = 1; i < argc; i++) {
-				if (!(fp = fopen(argv[i], "r")))
-					err(1, "fopen: %s", argv[i]);
+
+		for (i = 1; i < argc; i++) {
+			if (!(fp = fopen(argv[i], "r")))
+				err(1, "fopen: %s", argv[i]);
+			if (fstat(fileno(fp), &st) == -1)
+				err(1, "fstat: %s", argv[i]);
+
+			/* did the file change? by size, modification */
+			if (stfiles[i - 1].st_size != st.st_size ||
+			    stfiles[i - 1].st_mtime != st.st_mtime) {
 				name = ((name = strrchr(argv[i], '/'))) ? name + 1 : argv[i];
 				printfeed(fp, name);
 				if (ferror(fp))
 					err(1, "ferror: %s", argv[i]);
-				fclose(fp);
 			}
+			memcpy(&stfiles[i - 1], &st, sizeof(st));
+			fclose(fp);
 		}
-		/* DEBUG: TODO: gc first run. */
-		gc();
-
-		sleep(sleepsecs);
-		slept += sleepsecs;
 
-		/* gc once every hour (excluding run-time) */
-		if (slept >= 3600) {
+		/* "garbage collect" on a change or every 5 minutes */
+		if (changed || slept > 10) {
 			gc();
+			changed = 0;
 			slept = 0;
 		}
+		sleep(1);
+		slept++;
 	}
 	return 0;
 }