PHP filesystem函数和include require的性能研究

发布时间:August 5, 2007 分类:PHP

写在Magike Blog 1.0 Release发布以后



另外说句 其实有些时候还是google好用

It is not often that you can write a PHP script that does not need to include the contents of different files as part of it's output. If these includes happen to be php scripts themselves you have no choice but to use require or include.
However more often than not, the contents are static, usually html template component. With static includes you have many more options available.
We will analyse some of these functions to find out which one is most suitable when dealing with files with static content. We use the term function loosely, because require and include are not real functions but language constructs.

Function Brief Description
string file_get_contents ( string filename [, int use_include_path])Reads entire file into a string
int fpassthru ( resource handle)Output all remaining data on a file pointer
string fgets ( resource handle [, int length])Gets line from file pointer
array file ( string filename [, int use_include_path])Reads entire file into an array
require(string filename)
include(string filename)
require_once(string filename)
include_once(string filename)
includes and evaluates the specific file.
int readfile ( string filename [, int use_include_path])Outputs a file

We will now attempt to 'include' the contents of a 1 megabyte file into the output produced by our php script. How you can generate files of specific sizes is described elsewhere. The execution times and peak memory consumption, as reported by xdebug have been tabulated below.

We compensate for file caching and background processes by executing each script 4 times and taking the average (mean) of the result number 2-4. The first result is always rejected. Any result that appears to be outlier is rejected. The mean is rounded to 5 decimal places.

Function Sample Usage Time (s) Memory (b)
file_get_contentsecho file_get_contents($filename);0.005641067856
$fp = fopen($filename,"rb");
echo fgets($fp);
fileecho join("",file($filename));0.064642185624

What's obvious from these results is that using fpassthru is far superior to all other methods. What's not so obvious is that fpassthru and readfile are equally good. The fpassthru version runs 0.00007 seconds quicker than the readfile version. What that really means is that you need to run the script at least 100000 times to make significant saving. On memory consumption readfile seems to have use up around 1kb less than passthru. A kilo byte is a drop in the ocean for modern web servers with hundreds of megabytes if not gigabytes of memory.

The only conclusion that can be drawn from these studies is that fpassthru and readfile are equally good if you wish to include static content as part of the script's output.

Before you rush off to change all your includes and requires into readfiles or fpassthrus let's run the same test with a smaller (32Kb file). 32Kb is a more realistic size for an included file.

Function Time (s) Memory (b)
32Kb File1Mb File 32Kb File1Mb File
file_get_contents 0.001520.00564 524801067856
fpassthru 0.001170.00184 2001620032
fgets 0.001950.07190 3076030768
file 0.001570.06464 873442185624
require_once 0.002250.08065 679922067696
include 0.002220.08202 679282067624
readfile 0.001170.00191 1919219208

readfile and fpassthru have once again tied for first place. This new set of results just confirms the fact that speed and scalability comes from your design and not from your code. The difference between the best performance and the worst is just 0.00108s too close to call.

The most significant feature of these results is that both fpassthru and readfile scale really well. In other words, memory consumption and execution time does not increase significantly with increase in file size. That does not always mean your script will be faster just because you use these functions instead of require or include.

Tags: none

已有 5 条评论 »

  1. simophin simophin


    1. kimi kimi

      @simophin, 缓存数据是个很深的水,从开始以为的文本缓存,到内存缓存,再到中介缓存。太多的需要研究了。

  2. 笨蛋 笨蛋

    include = readfile+ (init valiable +calc )

    如果你仅仅缓存文本的 而不需要从文本中 初始化出变量.

  3. 杭州人流医院 杭州人流医院


  4. 杭州男科医院 杭州男科医院

    博主&笨蛋讲的有道理 学习下:)

添加新评论 »