(PHP 5, PHP 7, PHP 8)
stream_get_line — 从资源流里读取一行直到给定的定界符
$handle
, int $length
, string $ending = ?
) : string从给定的资源流里读取一行。
当读取到 length 个字节数就结束，或者当在读取的字符串中发现 ending （不包含到返回值里）也结束，又或者遇到了 EOF 也结束（总之以上条件中哪个先出现就以哪个为准）。
这个函数与 fgets() 几乎是相同的，唯一的区别是在这个函数里面允许指定行尾的定界符，而不是使用标准的 \n， \r 还有 \r\n ，并且返回值中不包含定界符。（翻译注：也可以把 \n 等作为定界符传入 ending ）
handle一个有效的文件句柄。
length需要从句柄里读取的字节数。
ending可选参数，字符串定界符。
返回一个字符串，该字符串的内容根据 length 字节数从 handle 里读取。
如果发生错误，则返回 false.
I have been struggling with the problem that stream_get_line() sometimes reads too much when the 3rd parameter is used (and the 3rd parameter has a length greater than 1), so $ending is actually contained in the data returned (https://bugs.php.net/bug.php?id=63240).
I don't have the option of upgrading my PHP version but it seems that a workaround can be to insert:
fseek($fp, ftell($fp));
just before calling stream_get_line().
If you are specifying the 3rd optional "ending" parameter as a string which is more than one character and actually find that the line returned by the function sometimes contains this "ending" value it may be related to the following bug, https://bugs.php.net/bug.php?id=63240
Our server was running 5.3.18 and when we upgraded to 5.3.20 it worked fine, I believe this was fixed in 5.3.19 though.
WARNING:
Specifying a length of 0 does NOT give you an infinite length, contrary to what the documentation might suggest. Instead, setting a length of 0 just makes the function default to a length of 8192. To be precise, it gets the value PHP_SOCK_CHUNK_SIZE (8192) in ext/standard/streamsfuncs.c.
So, let's say you're trying to read ALL data until you reach a "\x03" (decimal 3) byte. How do you GUARANTEE that this is the case? Well, there's no way! The only thing you can do stream_get_lin() into a "master" string, then fseek() backwards by 1 character, then fgetc() and verify that it's "\x03". If you don't see a "\x03", it means that stream_get_line() has aborted after 8192 bytes and before hitting "\x03", and you'll have to call it again. Keep appending the return values to a "master" string until you hit a "\x03" or EOF... That's the ONLY way to properly build a string that contains EVERYTHING until the character you're looking for.
fgets is faster but stream_get_line is more useful in a tcp server scripts.
when fgets reads some bytes from socket, where EOF is reached, it returns bool(false) same as stream_get_line
BUT if remote client drops connection, and server script will try to read some data with function fgets, function will return bool(false), and stream_get_line will return string(0) ""
so you can detect remote client disconnection with stream_get_line, and cannot with fgets
If the "ending" is a string, there are cases where the function doesn't return the correct value for the first time it is called. Don't be shocked if you find it returning a string value of upto "length" that includes the "ending". (See bug #44607)
If the "ending" is just a single character, the function would always work correctly. ("\n" is a single character)
Temporarily, until this is fixed, the below function can be used:
<?php
function istream_get_line(&$fp, $length, $end) {
$current = ftell($fp);
$str = fread($fp, $length);
$i = strpos($str, $end);
if ($i === FALSE) {
return $str;
} else {
fseek($fp, $current + $i + strlen($end));
return substr($str, 0, $i);
}
}
?>
I've spent quite a while trying to get stream_get_line to get a chunk encoded html file and to finish correctly at the end so that I can pipeline requests.
This is the function I have come up with.
<?php
function getURLContents($url, $ip, $port, $ssl = false, $closeConnection = false)
{
if ($ssl)
$ssl = 'ssl://';
else
$ssl = '';
$fp = pfsockopen($ssl.$ip, $port, $errno, $errstr, MAX_TIME_TO_START_CONNECTION);
if ($fp)
{
$out = 'GET '.$url." HTTP/1.1\r\n";
$out .= 'Host: '.$ip.':'.$port."\r\n";
if ($closeConnection)
$out .= "Connection: close\r\n";
else
$out .= "Connection: keep-alive\r\n";
$out .= "\r\n";
if (!fwrite($fp, $out))
{
echo 'Problem writing to socket, opening a new connection.';
fclose($fp);
$fp = pfsockopen($ssl.$ip, $port, $errno, $errstr, MAX_TIME_TO_START_CONNECTION);
fwrite($fp, $out);
}
$theData = '';
$notDone = true;
stream_set_blocking($fp, 0);
$startTime = time();
$lastTime = $startTime;
while (!feof($fp) && !$done && (($startTime + MAX_TIME_FOR_THE_RESPONSE) > time()))
{
usleep(100);
$theNewData = stream_get_line($fp, 1024, "\n");
$theData .= $theNewData;
$done = (trim($theNewData) === '0');
}
}
else
{
echo 'ERROR CONNECTING TO '.$ip.':'.$port;
return false;
}
if ($closeConnection)
fclose($fp);
return $theData;
}
?>
My testing has found this function to be dramatically faster than fgets on PHP 5.1.14. The difference is probably due to how buffering is used internally. Compare the following:
<?php
// reads 10,000 lines in 27 seconds
while (!feof($handle)) {
$line = fgets($handle, 1000000);
}
?>
vs.
<?php
// reads 10,000 lines in 0.5 seconds
while (!feof($handle)) {
$line = stream_get_line($handle, 1000000, "\n");
}
?>
In version 5.0.4 using this funtion and then calling ftell($stream) would give you the position up to but not including the "ending" string.
When I rev'd to PHP version 5.1.2, calling this function then using ftell($stream) would give the position up to AND including the "ending" string
for example, parsing HTTP responses.
The response from apache using curl....
------------------------------------------------------------
HTTP/1.1 200 OK
Date: Tue, 18 Apr 2006 20:54:59 GMT
Server: Apache/1.3.33 (Unix) PHP/5.0.4 mod_ssl/2.8.22 OpenSSL/0.9.7e
X-Powered-By: PHP/5.0.4
Transfer-Encoding: chunked
Content-Type: text/html
<html><body>test</body></html>
-------------------------------------------------------------
The code:
<?php
$headers = stream_get_line($in,4096,"\r\n\r\n");
fseek ($in,ftell($in)+4);
while (!feof($in)){
fputs ($out,stream_get_line($in,4096,''));
}
?>
prior to my 5.0.4 this worked perfectly, trimming the \r\n\r\n section of the HTTP response and seperating the top into the $headers string, and the rest was placed into the file handle $out.
using php 5.1.2, the above code chopps off the first 4 bytes of the HTTP response and puts
l><body>test</body></html>
into $out.