Python file.read() 实际上获取了多余的数据

2 投票
2 回答
515 浏览
提问于 2025-04-16 01:02
cat file_ro.py 
import sys
def file_open(filename):
        fo=open(filename,'r')
        fo.seek(7)
        read_data=fo.read(3)
        fo.close()
        print read_data
file_open("file.py")

但是strace显示

readlink("file_ro.py", 0x7fff31fc7ea0, 4096) = -1 EINVAL (Invalid argument)
getcwd("/home/laks/python", 4096)       = 18
lstat("/home/laks/python/file_ro.py", {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
stat("file_ro.py", {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
open("file_ro.py", O_RDONLY)            = 3
fstat(3, {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa671a6c000
fstat(3, {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
lseek(3, 0, SEEK_SET)                   = 0
read(3, "import sys\ndef file_open(filenam"..., 128) = 128
read(3, "ile_open(\"file.py\")\n\t\n", 4096) = 22
close(3)                                = 0
munmap(0x7fa671a6c000, 4096)            = 0
stat("file_ro.py", {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
open("file_ro.py", O_RDONLY)            = 3
fstat(3, {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
ioctl(3, SNDCTL_TMR_TIMEBASE or TCGETS, 0x7fff31fc9e30) = -1 ENOTTY (Inappropriate ioctl for device)
fstat(3, {st_mode=S_IFREG|0755, st_size=150, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa671a6c000
lseek(3, 0, SEEK_CUR)                   = 0
read(3, "import sys\ndef file_open(filenam"..., 4096) = 150
lseek(3, 150, SEEK_SET)                 = 150
read(3, "", 4096)                       = 0
close(3)                                = 0
munmap(0x7fa671a6c000, 4096)            = 0
open("file.py", O_RDONLY)               = 3
fstat(3, {st_mode=S_IFREG|0755, st_size=305, ...}) = 0
fstat(3, {st_mode=S_IFREG|0755, st_size=305, ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa671a6c000
lseek(3, 0, SEEK_SET)                   = 0
read(3, "import ", 7)                   = 7
read(3, "sys\ndef file_open(filename):\n\t\"\""..., 4096) = 298
close(3)                                = 0
munmap(0x7fa671a6c000, 4096)            = 0
fstat(1, {st_mode=S_IFCHR|0620, st_rdev=makedev(136, 0), ...}) = 0
mmap(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0x7fa671a6c000
write(1, "sys\n", 4sys
)                    = 4
rt_sigaction(SIGINT, {SIG_DFL, [], SA_RESTORER, 0x306140efa0}, {0x306d10b2b0, [], SA_RESTORER, 0x306140efa0}, 8) = 0
close(5)                                = 0
munmap(0x7fa671952000, 4096)            = 0
exit_group(0)  

你可以看到上面 -

read(3, "import sys\ndef file_open(filenam"..., 4096) = 150

为什么read()函数返回了150个字节,而程序只要求读取3个字节呢?

2 个回答

6

缓冲。为了避免这个问题,可以使用 open(filename, 'rb', bufsize=0)

6

因为你是在另一个.py文件中读取东西,所以可能会让事情变得有些混乱。不过,看起来内置的读取函数会忽略你传给read()的值,而是把剩下的值缓存起来。也许你可以试试用os.read()来代替?

file_ro.py:

import sys
def file_open(filename):
        fo=open(filename,'r')
        fo.seek(7)
        read_data=fo.read(3)
        fo.close()
        print read_data
file_open("zzz")

zzz:

12345678901234567890123456789012345678901234567890

strace:

...
open("zzz", O_RDONLY|O_LARGEFILE)       = 3
fstat64(3, {st_mode=S_IFREG|0644, st_size=51, ...}) = 0
fstat64(3, {st_mode=S_IFREG|0644, st_size=51, ...}) = 0
mmap2(NULL, 4096, PROT_READ|PROT_WRITE, MAP_PRIVATE|MAP_ANONYMOUS, -1, 0) = 0xb73fb000
_llseek(3, 0, [0], SEEK_SET)            = 0
read(3, "1234567", 7)                   = 7
read(3, "89012345678901234567890123456789"..., 4096) = 44
close(3)                                = 0
...

你可以在打开文件时指定缓存的大小,比如用open('zzz', buffering=0),或者我使用了os模块,这样可以更好地控制文件的读取,正如你想要的那样:

file_ro2.py:

import sys, os
def file_open(filename):
        fo=os.open(filename, os.O_RDONLY)
        os.lseek(fo, 7, 0)
        read_data=os.read(fo, 3)
        os.close(fo)
        print read_data
file_open("zzz")

strace2:

...
open("zzz", O_RDONLY|O_LARGEFILE)       = 3
_llseek(3, 7, [7], SEEK_SET)            = 0
read(3, "890", 3)                       = 3
close(3)                                = 0
...

撰写回答