Bash：用文件名中的空格扩展大括号和全局？
声明：本页面是StackOverFlow热门问题的中英对照翻译，遵循CC BY-SA 4.0协议，如果您需要使用它，必须同样遵循CC BY-SA许可，注明原文地址和作者信息，同时你必须将它归于原作者（不是我）：StackOverFlow
原文地址: http://stackoverflow.com/questions/14381711/
Warning: these are provided under cc-by-sa 4.0 license. You are free to use/share it, But you must attribute it to the original authors (not me):
StackOverFlow
Bash: Expand braces and globs with spaces in filenames?
提问by Jay Hacker
I have some files that look like:
我有一些看起来像的文件：
/path/with spaces/{a,b,c}/*.gz
And I need all files matching the glob under a subset of the a,b,cdirs to end up as arguments to a single command:
我需要在a,b,c目录的子集下与 glob 匹配的所有文件最终作为单个命令的参数：
mycmd '/path/with spaces/a/1.gz' '/path/with spaces/a/2.gz' '/path/with spaces/c/3.gz' ...
The directories I care about come in as command line params and I have them in an array:
我关心的目录作为命令行参数进来，我把它们放在一个数组中：
dirs=( "$@" )
And I want to do something like:
我想做类似的事情：
IFS=,
mycmd "/path/with spaces/{${dirs[*]}}/"*.gz
but this doesn't work, because bash expands braces before variables. I have tried tricks with echoand lsand even eval(*shudder*) but it's tough to make them work with spaces in filenames. finddoesn't seem to be much help because it doesn't do braces. I can get a separate glob for each dir in an array with:
但这不起作用，因为 bash 在变量之前扩展大括号。我已经尝试过使用echoandls甚至eval(*shudder*) 的技巧，但很难让它们与文件名中的空格一起使用。 find似乎没有多大帮助，因为它不做大括号。我可以为数组中的每个目录获取一个单独的 glob：
dirs=( "${dirs[@]/#//path/with spaces/}" )
dirs=( "${dirs[@]/%//*.gz}" )
but then bash quotes the wildcards on expansion.
但随后 bash 在扩展时引用通配符。
So: is there an elegant way to get all the files matching a variable brace and glob pattern, properly handling spaces, or am I stuck doing for loops? I'm using Bash 3 if that makes a difference.
那么：是否有一种优雅的方法可以让所有文件与变量大括号和 glob 模式匹配，正确处理空格，还是我一直在做 for 循环？如果这有所作为，我将使用 Bash 3。
采纳答案by Jay Hacker
回答by Shawn Chin
To perform brace expansion and globbing on a path with spaces, you can quote the portions of the path that contain spaces, e.g.
要在带有空格的路径上执行括号扩展和通配符，您可以引用路径中包含空格的部分，例如
mycmd '/path/with spaces/'{a,b,c}/*.gz
Doing brace expansion using a list of values from a variable is a little tricky since brace expansion is done before any other expansion. I don't see any way but to use the dreaded eval.
使用来自变量的值列表进行大括号扩展有点棘手，因为大括号扩展是在任何其他扩展之前完成的。除了使用可怕的eval.
eval mycmd "'/path/with spaces/'{a,b,c}/*.gz"
P.S. In such a case however, I would personally opt for a loop to build the argument list rather than the approach shown above. While more verbose, a loop will be a lot easier to read for the uninitiated and will avoid the need to use eval(especially when the expansion candidates are derived from user input!).
PS 但是，在这种情况下，我个人会选择循环来构建参数列表，而不是上面显示的方法。虽然更冗长，但对于未入门的人来说，循环会更容易阅读，并且无需使用eval（尤其是当扩展候选来自用户输入时！）。
Proof of concept:
概念证明：
Using a dummy command (x.sh) which prints out the number of arguments and prints out each argument:
使用打印出参数数量并打印出每个参数的虚拟命令 (x.sh)：
[me@home]$ shopt -s nullglob # handle case where globbing returns no match
[me@home]$ ./x.sh 'path with space'/{a,b}/*.txt
Number of arguments = 3
- path with space/a/1.txt
- path with space/b/2.txt
- path with space/b/3.txt
[me@home]:~/temp$ dirs="a,b"
[me@home]k:~/temp$ eval ./x.sh "'path with space'/{$dirs}/*.txt"
Number of arguments = 3
- path with space/a/1.txt
- path with space/b/2.txt
- path with space/b/3.txt
回答by Jay Hacker
Here's one for the GNU Parallelfans:
这是GNU Parallel粉丝的一个：
parallel -Xj1 mycmd {}/*.gz ::: "${dirs[@]/#//path/with spaces/}"
