# 运算符优先级

[array()

* / % 算数运算符
+ - . 算数运算符字符串运算符
<< >> 位运算符

& 位运算符引用
^ 位运算符
| 位运算符
&& 逻辑运算符
|| 逻辑运算符
? : 三元运算符
= += -= *= /= .= %= &= |= ^= <<= >>= 赋值运算符
and 逻辑运算符
xor 逻辑运算符
or 逻辑运算符
,多处用到

 ``` \$a = 5, \$b = 5?> ```

User Contributed Notes
01-Nov-2006 02:20
``` In response to: ---------------------- T Chan 19-Aug-2006 10:30 [...] In response to the manual, associativity doesn't determine the order of evaluation - it determines how it's "bracketed": a=b=c=1 is equivalent to a=(b=(c=1)), but a is evaluated first (an important point if evaluating a has side-effects) <?php function one(\$str) {    echo "\$str";    return 1; } \$a = array(1,2); \$a[one("A")] = \$a[one("B")] = 1; // outputs AB ?> ---------------------- The manual, stated "associativity determines order of evaluation," which is completely correct, as it uses a stack execution, as with C. The example you provided to counter the manual's statement however, is contrived, as it uses function calls to output a variable which is not even used in assignment; this in turn disproves (in your opinion) the manual's statement. Fuction calls, have much higher precedence than the assignment operator. However: Right Associativity with the '=' operator means the following: The expression:  \$a = \$b = 1; IS indeed evaluated from right to left. The stack is used as follows. (if you're not familiar with stacks / queues, look em up). The expressions are pushed onto the stack in the following order:         \$a    \$b  =    1    =        ^                        ^        Bottom                Top The stack is then "evaluated" from top to bottom (which translates from right to left, if you're reading the expression) It would read:  assign 1 to assign \$b to \$a Which results in \$a = 1 Try the following simple script: <?php function output(& \$b) {      echo( "b value is: ".\$b );    return \$b; }  \$a = 2;  \$b = 5;  echo (\$a = \$b = 1);  \$a = 2;  \$b = 5;  echo ( "a value is: ".\$a = output(\$b) ); ?> Output is: 1 b value is: 5 a value is: 5 --------------------------------------- Your example had nothing to do with the assignment operator evaluation order, only procedural precedence. ```
T Chan
20-Aug-2006 01:30
``` In response to 26-Mar-2001 08:53, parens don't need to have precedence. There's only one way to convert them to a syntax tree. I can't think of a sensible reason it could be ambiguous either (and sensible language designers will make ( always pair with ) so there can't be any ambiguity). In response to 12-Aug-2005 08:47, you can do <?php \$myvar OR (\$myvar = 1); ?>, or the equivalent <?php !\$myvar AND \$myvar = 1; ?>, and if you have to use ifs, <?php if(!\$myvar) { \$myvar = 1; } ?> will do just fine (though in general, I'd be careful using boolean expressions as an indicator of success/failure when 0 could be a valid value. And you probably want to use isset(\$myvariable)). In response to the manual, associativity doesn't determine the order of evaluation - it determines how it's "bracketed": a=b=c=1 is equivalent to a=(b=(c=1)), but a is evaluated first (an important point if evaluating a has side-effects) <?php function one(\$str) {    echo "\$str";    return 1; } \$a = array(1,2); \$a[one("A")] = \$a[one("B")] = 1; // outputs AB ?> ```
golotyuk at gmail dot com
10-Jul-2006 12:51
``` Simple POST and PRE incremnt sample: <?php \$b = 5; \$a = ( ( ++\$b ) > 5 ); // Pre-increment test echo (int)\$a; \$b = 5; \$a = ( ( \$b++ ) > 5 ); // Post-increment test echo (int)\$a; ?> This will output 10, because of the difference in post- and pre-increment operations ```
dlyons at lyons42 dot com
26-Nov-2005 07:30
``` Re: Rick on 2-Sep-2005. Actually, the C equivalent of "\$a[\$c++]=\$b[\$c++];" has undefined behavior, and the increments are by no means guaranteed to happen after the assignment in C.  (Search for a discussion of C "sequence points" for details.) ```
sm
02-Sep-2005 08:15
``` Note the highly unfortunate difference from Java, which associates the trinary operator right-to-left. ---------------------------- source function trinaryTest(\$foo){  // works as you think in Java, but not PHP    \$bar    = \$foo > 20            ? "greater than 20"            : \$foo > 10                ? "greater than 10"                : \$foo > 5                    ? "greater than 5"                    : "not worthy of consideration";       echo \$foo." =>  ".\$bar."\n"; } echo "\n\n\n----trinaryTest\n\n"; trinaryTest(21); trinaryTest(11); trinaryTest(6); trinaryTest(4); function trinaryTestParens(\$foo){       \$bar    = \$foo > 20            ? "greater than 20"            : (\$foo > 10                ? "greater than 10"                : (\$foo > 5                    ? "greater than 5"                    : "not worthy of consideration"));       echo \$foo." =>  ".\$bar."\n"; } echo "\n\n\n----trinaryTestParens\n\n"; trinaryTestParens(21); trinaryTestParens(11); trinaryTest(6); trinaryTestParens(4); ---------------------------- output ----trinaryTest 21 =>  greater than 5 11 =>  greater than 5 6 =>  greater than 5 4 =>  not worthy of consideration ----trinaryTestParens 21 =>  greater than 20 11 =>  greater than 10 6 =>  greater than 5 4 =>  not worthy of consideration ```
rick at nomorespam dot fourfront dot ltd dot uk
02-Sep-2005 06:51
``` A quick note to any C developers out there, assignment expressions are not interpreted as you may expect - take the following code ;- <?php \$a=array(1,2,3); \$b=array(4,5,6); \$c=1; \$a[\$c++]=\$b[\$c++]; print_r( \$a ) ; ?> This will output;- Array ( [0] => 1 [1] => 6 [2] => 3 ) as if the code said;- \$a[1]=\$b[2]; Under a C compiler the result is;- Array ( [0] => 1 [1] => 5 [2] => 3 ) as if the code said;- \$a[1]=\$b[1]; It would appear that in php the increment in the left side of the assignment is processed prior to processing the right side of the assignment, whereas in C, neither increment occurs until after the assignment. ```
kit dot lester at lycos dot co dot uk
22-Aug-2005 08:38
``` D'oh! please ignore (& forgive) the first paragraph in my note yesterday - I said that a diadic operator had the same precedence & associativity as a bunch of monadics. A bit of early senility must have struck me. When I find time I'll rework my test to find out what I should have said - but using instanceof monadically didn't cause any visible errors in the test, so it could take a while. [I'd be delighted if someone else beats me to it/ spares me the difficulties of seeing what's wrong in something that shouldn't have worked.] ```
kit dot lester at lycos do co dot uk
22-Aug-2005 12:21
``` Table 15-1 omits the precedence of instanceof - testing suggests it to be of the same precedence as not, negate, casting, and @. The table also omits the precedence (and associativity) of the "execution operator" - but since that's a sort of quoting, I don't think it meaningfully has a precedence or associativity - they explain what is to happen where there's an open-endedness in the sense of missing brackets, and the quoting is a sort of bracket. (At least: because of the execution operator's double-ended closedness, I can't figure out any code where it would matter, so I can't test it.) ```
webmaster AT cafe-clope DOT net
12-Aug-2005 03:47
``` I regularly use some syntax like : <?php if(!\$myvar)   \$myvar = \$value ; ?> and <?php if(\$myvar)   echo "myvar is \$myvar today" ; ?> (or <?php echo (\$myvar ? "myvar is \$myvar today" : "") ?>) It's small, but can become heavy when used too much. Isn't there some trick to better such syntaxes ? I was wondering about using things like : <?php \$myvar ||= \$value ; ?> <?php \$myvar ?= \$value ; ?> <?php echo (\$myvar ? "myvar is \$myvar today") ; ?> ```
edwardsbc at yahoo dot com
05-May-2005 01:26
``` In response to npeelman at cfl dot rr dot com 29-Dec-2004 06:22: You have misunderstood the behaviour of the interpreter. With out the curly braces and the single quoted key identifier, the interpreter "assumes" you meant your CONSTANT to be a string.  This ONLY works within a parsed (double quoted) string.  And it doesn't help you at all if your array is multi-dimensional. I consider this a very bad habbit because it will get you in trouble elsewhere. Try the following: <?php define('item','AnyOldThing'); define('b',12); \$arr['item']['b'] = 'string'; \$arr['AnyOldThing'][12]= 'Maybe not what I intended.'; echo "This is a {\$arr['item']['b']}"; // [1] prints "This is a string". echo "This is a \$arr[item][b]"; // [2] broken echo \$arr[item][b]; // [3] broken ?> ```
npeelman at cfl dot rr dot com
29-Dec-2004 07:22
``` Update to message by yasuo_ohgaki at hotmail dot com:   I know this is an old message but when using an Associative array in a string you do not have to use { and  } to resolve ambiguity. ex: Associative Array in string: \$arr['item'] = 'string'; echo "This is {\$arr['item']}"; //prints "This is string". ...does work but, so does: echo "This is \$arr[item]"; //prints "This is string". ... simply enclose the whole string with double quotes and leave out the single quotes from around the index name. This simplifies the code and makes things easier to read. ```
Stopping at the dot completely
01-Sep-2004 04:33
``` The low precedence of the OR operator is useful for error control statements such as this one: \$my_file = @file ('non_existent_file') or die ("Failed opening file: error was '\$php_errormsg'"); Notice the good readability of the code. ```
23-Aug-2004 12:51
``` I think warhog's note about the differing precedence between && / AND and || / OR is worth repeating.  Since && and || evaluate before the assignment operator (=) while AND and OR evaluate after it, you can get COMPLETELY different results if you don't fully parenthesise. I cannot imagine when it would ever be important that those two pairs have differing precedence, but they do.  And I just spent two hours discovering that the hard way because I broke my career-long rule: *Always fully parenthesise!* ```
11-Jun-2004 10:22
``` Warhog wrote: "maybe usefull for some tricky coding and helpfull to prevent bugs :D" I'm sure Warhog was being facetious, but for the new programmers in the audience I'd like to point out that 'tricky coding' and relying on precedence/order of evaluation are both well-known ways to *produce* bugs. Use parentheses instead. ```
10-Jun-2004 08:58
``` of course this should be clear, but i think it has to be mentioned espacially: AND is not the same like && for example: <?php \$a && \$b || \$c; ?> is not the same like <?php \$a AND \$b || \$c; ?> the first thing is (a and b) or c the second a and (b or c) 'cause || has got a higher priority than and, but less than && of course, using always [ && and || ] or [ AND and OR ] would be okay, but than you should at least respect the following: <?php \$a = \$b && \$c; ?> <?php \$a = \$b AND \$c; ?> the first code will set \$a to the result of the comparison \$b with \$c, both have to be true, while the second code line will set \$a like \$b and THAN - after that - compare the success of this with the value of \$c maybe usefull for some tricky coding and helpfull to prevent bugs :D greetz, Warhog ```
yasuo_ohgaki at hotmail dot com
26-Mar-2001 04:34
``` About "{" and "}". Sometimes PHP programmers need to use "{" and "}" to resolve ambiguity. Here is some examples. Variable Variables: \$foo = "test"; \$\$bar = "this is"; echo "\${\$bar} \$foo"; // prints "this is test" Note: it is the same as echo "\$test \$foo"; Array in string: \$arr[10][10][10] = "string"; echo "This is {\$arr[10][10][10]}"; // prints "This is string" Associative Array in string: \$arr['item'] = 'string'; echo "This is {\$arr['item']}"; //prints "This is string". ```
yasuo_ohgaki at hotmail dot com
26-Mar-2001 03:53
``` Other Language books' operator precedence section usually include "(" and ")" - with exception of a Perl book that I have. (In PHP "{" and "}" should also be considered also). However, PHP Manual is not listed "(" and ")" in precedence list. It looks like "(" and ")" has higher precedence as it should be. Note: If you write following code, you would need "()" to get expected value. \$bar = true; \$str = "TEST". (\$bar ? 'true' : 'false') ."TEST"; Without "(" and ")" you will get only "true" in \$str. (PHP4.0.4pl1/Apache DSO/Linux, PHP4.0.5RC1/Apache DSO/W2K Server) It's due to precedence, probably. ```