Hmm so, I took my own advice and psychon's and split it into 3
patches, with the last being kind of unnecessary but I'd love to see
it accepted anyway :>

In attached patch application order:
1) 0001-Deduplicate-the-warn-in-in-color_init_unchecked.patch

This patch gets rid of the same warn() message being sent in
color_init_unchecked().  The problem is color_parse() sends the same
message and unchecked() sends if color_parse() fails.  The duplicated
message is somewhat unhelpful.

2) 0001-Use-RGB_16TO8-in-luaA_push_color.patch

This patch makes use of RGB_16TO8() in luaA_push_color() which was
defined but not used.  I also attempted to clean up how the
RGB_16TO8() and RGB_8TO16() macros were written.  The [good] compiler
would of course equate 0xFFFF / 0xFF to 0x101 as a constant expression
but I thought it was clearer written my way with a comment to explain
what 0x101 is. I think I'll also keep a lookout for anywhere else in
awesome that might have missed the use of these RGB_ macros.  The main
thing is the change to luaA_push_color() relying on the macros rather
than duplicated code.

3) 0001-Imperceptable-optimization-to-color_parse.patch

This patch is unnecessary but I like how I wrote it rather than the
original. :> I made a minor change to the if statement that warn()'s
on an invalid color so it *might* short-circuit earlier and then I
also rewrote the section where the first 3 octets of colnum were split
into their own uint8_t red, green, and blue values.  I know *red =
(colnum >> 16) & 0xff; is pretty clear already but if it were up to me
I would shift it over octet by octet to make it transparent to the
skimming reader than the colors are similarly being popped off of an
array.  The first time I saw the (colnum >> 16) portion I wasn't quite
sure where red was, it being the last octet masked off of colnum seems
clearer to me that it is the third octet being harvested outright.
Furthermore (colnum >> 16) on a good but not perfect compiler creates
a temporary, (colnum >> 8) does also.  My way reuses colnum which is
hopefully already in a register.  colnum isn't used after that at all
so the modification of colnum is perfectly fine for its current use.
:>  I have been told by a friend that a *great* compiler would notice
that colnum isn't used at all after this and would instead modify
colnum without creating temporaries from it.  But I don't trust
compilers to be that omnicient.  Still, if this means nothing then I'm
disappointed but I won't start a holy war if this patch isn't
accepted.

Isn't it just ever so much fun to obsess about the little things? :p

Toodles :]

Patch order:
1) 0001-Deduplicate-the-warn-in-in-color_init_unchecked.patch
2) 0001-Use-RGB_16TO8-in-luaA_push_color.patch
3) 0001-Imperceptable-optimization-to-color_parse.patch

On Fri, Nov 18, 2011 at 10:37 AM, Majic <[email protected]> wrote:
> So I have a few minor changes to color.c :>
>
> I actually posted this linked from sprunge.us a couple times in
> #awesome and it wasn't correct all those times.  So this is meant to
> be the final patch where everything is good and happy. :x
>
> See attached.
>
> Would love some peer review and maybe eventually accepted into master?
>  The only validly good thing about this patch is it removes a problem
> where if color_parse() failed it would warn about the color being
> invalid twice.  (Since the warn() was sent again after the return of
> color_parse() in color_init_unchecked()  Maybe this should be two
> patches, or three, but I felt that all of it was simple enough to be
> one :>  If it were 3 it would be "deduplicate warn() used in
> color_parse() and color_init_unchecked()", make the bit harvesting in
> color_parse() (possibly?) more understandable and combine the if's in
> color_init_unchecked()", then "rework the RGB_8TO16/16TO8 macros and
> use 16TO8 in luaA_push_color()"  But that's like 3 patches for pretty
> simple changes. :>
>
> The important thing to remember is 0xFFFF / 0xFF == 0x101 >.<
>
> Toodles.
>
> PS: The only reason I touched the if in color_parse() was to help it
> short-circuit sooner.
>
From 552ee90b1fcb38bff3c0313e689015c91f1fadf2 Mon Sep 17 00:00:00 2001
From: Majic <[email protected]>
Date: Fri, 18 Nov 2011 20:36:34 -0800
Subject: [PATCH] Deduplicate the warn() in in color_init_unchecked()

---
 color.c |   18 ++++--------------
 1 files changed, 4 insertions(+), 14 deletions(-)

diff --git a/color.c b/color.c
index ceee8d4..faba448 100644
--- a/color.c
+++ b/color.c
@@ -74,31 +74,21 @@ color_init_unchecked(color_t *color, const char *colstr, ssize_t len)
 
     p_clear(&req, 1);
 
-    if(!len)
-    {
-        req.has_error = true;
-        return req;
-    }
-
-    req.color = color;
-
     /* The color is given in RGB value */
-    if(!color_parse(colstr, len, &red, &green, &blue))
+    if(len == 0 || !color_parse(colstr, len, &red, &green, &blue))
     {
-        warn("awesome: error, invalid color '%s'", colstr);
         req.has_error = true;
         return req;
     }
 
+    req.color       = color;
+    req.has_error   = false;
+    req.colstr      = colstr;
     req.cookie_hexa = xcb_alloc_color_unchecked(globalconf.connection,
                                                 globalconf.default_cmap,
                                                 RGB_8TO16(red),
                                                 RGB_8TO16(green),
                                                 RGB_8TO16(blue));
-
-    req.has_error = false;
-    req.colstr = colstr;
-
     return req;
 }
 
-- 
1.7.7.3

From d68e17e5c11255ff9d09d79bddee6839a1313b25 Mon Sep 17 00:00:00 2001
From: Majic <[email protected]>
Date: Fri, 18 Nov 2011 20:46:39 -0800
Subject: [PATCH] Use RGB_16TO8() in luaA_push_color()

---
 color.c |   11 ++++++-----
 1 files changed, 6 insertions(+), 5 deletions(-)

diff --git a/color.c b/color.c
index faba448..08df0dc 100644
--- a/color.c
+++ b/color.c
@@ -26,8 +26,9 @@
 #include "globalconf.h"
 #include "common/xutil.h"
 
-#define RGB_8TO16(i)   (0xffff * ((i) & 0xff) / 0xff)
-#define RGB_16TO8(i)   (0xff * ((i) & 0xffff) / 0xffff)
+/* 0xffff / 0xff == 0x101 */
+#define RGB_8T016(i) (((i) & 0xff)   * 0x101)
+#define RGB_16TO8(i) (((i) & 0xffff) / 0x101)
 
 /** Parse an hexadecimal color string to its component.
  * \param colstr The color string.
@@ -128,9 +129,9 @@ color_init_reply(color_init_request_t req)
 int
 luaA_pushcolor(lua_State *L, const color_t c)
 {
-    uint8_t r = (unsigned) c.red   * 0xff / 0xffff;
-    uint8_t g = (unsigned) c.green * 0xff / 0xffff;
-    uint8_t b = (unsigned) c.blue  * 0xff / 0xffff;
+    uint8_t r = RGB_16TO8(c.red);
+    uint8_t g = RGB_16TO8(c.green);
+    uint8_t b = RGB_16TO8(c.blue);
     char s[10];
     int len = snprintf(s, sizeof(s), "#%02x%02x%02x", r, g, b);
     lua_pushlstring(L, s, len);
-- 
1.7.7.3

From b4179f9493b08ff87082daaa22aa88ad8d50d221 Mon Sep 17 00:00:00 2001
From: Majic <[email protected]>
Date: Fri, 18 Nov 2011 20:50:20 -0800
Subject: [PATCH] Imperceptable optimization to color_parse()

---
 color.c |    8 +++++---
 1 files changed, 5 insertions(+), 3 deletions(-)

diff --git a/color.c b/color.c
index 08df0dc..cc363c1 100644
--- a/color.c
+++ b/color.c
@@ -46,15 +46,17 @@ color_parse(const char *colstr, ssize_t len,
     char *p;
 
     colnum = strtoul(colstr + 1, &p, 16);
-    if(len != 7 || (p - colstr) != 7 || colstr[0] != '#')
+    if(len != 7 || colstr[0] != '#' || (p - colstr) != 7)
     {
         warn("awesome: error, invalid color '%s'", colstr);
         return false;
     }
 
-    *red   = (colnum >> 16) & 0xff;
-    *green = (colnum >> 8) & 0xff;
     *blue  = colnum & 0xff;
+    colnum >>= 8;
+    *green = colnum & 0xff;
+    colnum >>= 8;
+    *red   = colnum & 0xff;
 
     return true;
 }
-- 
1.7.7.3

Reply via email to