ReaderUtil::each_byte includes trailing EOF

[evmar]

use io::ReaderUtil;

fn main() {
    let stdin = io::stdin();
    for stdin.each_byte() |b| {
        io::println(fmt!("byte '%c' %d", b as char, b));
    }
}

$ echo foo | ./main 
byte 'f' 102
byte 'o' 111
byte 'o' 111
byte '
' 10
byte '?' -1

That last byte shouldn't be included.

[evmar]

Using Rust 9143688 pulled yesterday.

[evmar]

Issue #2004 is probably related, in that it wasn't clear whether it's ok for read_byte to return -1.

[evmar]

Something like

From 7d649f37476f533de1bbd5c964350e7a9cc93ac3 Mon Sep 17 00:00:00 2001
From: Evan Martin <martine@danga.com>
Date: Wed, 20 Feb 2013 10:46:51 -0800
Subject: [PATCH] ReaderUtil::each_byte shouldn't include EOF byte -- Issue
 #5056

---
 src/libcore/io.rs | 18 ++++++++++++++++--
 1 file changed, 16 insertions(+), 2 deletions(-)

diff --git a/src/libcore/io.rs b/src/libcore/io.rs
index 2173efe..84877e7 100644
--- a/src/libcore/io.rs
+++ b/src/libcore/io.rs
@@ -277,12 +277,16 @@ impl<T: Reader> ReaderUtil for T {
     }

     fn each_byte(&self, it: fn(int) -> bool) {
-        while !self.eof() {
-            if !it(self.read_byte()) { break; }
+        loop {
+            match self.read_byte() {
+                -1 => break,
+                ch => if !it(ch) { break; }
+            }
         }
     }

     fn each_char(&self, it: fn(char) -> bool) {
+        // FIXME this doesn't handle EOF properly -- see issue #5056
         while !self.eof() {
             if !it(self.read_char()) { break; }
         }
@@ -1211,6 +1215,16 @@ mod tests {
     }

     #[test]
+    fn test_each_byte_empty() {
+        // Issue #5056 -- shouldn't include trailing EOF.
+        do io::with_str_reader(~"") |inp| {
+            for inp.each_byte() |b| {
+                assert(b != -1);
+            }
+        }
+    }
+
+    #[test]
     fn test_readchars_empty() {
         do io::with_str_reader(~"") |inp| {
             let res : ~[char] = inp.read_chars(128);
-- 
1.8.1.3

But the build is soooo slowwwwwww

[emberian]

@martine can you open a PR with that patch please?

[emberian]

(Reproduces)

[stepancheg]

Pull request #8081.

[evmar]

Sorry for never cleaning up my initial patch, I think I had imagined things would be thrown away and redone before my patch was ever useful. I think ultimately it's pretty busted to provide a .eof() method because of the issue mentioned in your patch -- you can't know if you're at the end of a file until you attempt to read from it, so you need a way for read to signal eof, at which point you don't so much need a second API that tells you whether subsequent reads will fail.

[stepancheg]

@martine I agree, .eof() is wrong operation for input stream. I'd drop it without replacement.

[stepancheg]

Pull request merged into master, issue can be closed now.