>> file command
Does it mean that all stdout is redirected, not only that from command?
Such redirection affects a simple command. From man 1 bash:
Before a command is executed, its input and output may be redirected using a special notation interpreted by the shell. […] The following redirection operators may precede or appear anywhere within a simple command or may follow a command. Redirections are processed in the order they appear, from left to right.
"The following redirection operators" are [n]<word, [n]>word, [n]>>word etc.
A simple command is a sequence of optional variable assignments followed by blank-separated words and redirections, and terminated by a control operator.
And
control operator
A token that performs a control function. It is one of the following symbols:
|| & && ; ;; ( ) | |& <newline>
This means the following commands are equivalent:
echo Some text >> file
echo Some >> file text
echo >> file Some text
>> file echo Some text
The question is tagged bash and I quote man 1 bash but the above commands work in sh as well.
The command line parser needs to "serve" all redirections before it runs the command (i.e. the command stripped of these redirections). Think about it, the procedure is the same regardless of where the particular redirection is. There's no reason to require it to be at the very end.
When and why would you reverse the order?
I don't recall a situation where I wanted to have a redirection in the middle. There is, however, a usage case where having an input redirection at the beginning is quite useful. Consider this code:
grep foo file | tail | …
Imagine the pipe is very long. It's posted on Super User and solves some problem. You may paste it to your console and run it. What if you'd rather want to append it to some command or pipe? E.g. you want to get:
my_custom_command | grep foo | tail | …
# ^^^^^^^^^^^^^^^^^^^ this is the part you'd be happy to paste
You need to remove file from the copied command. For this reason some users prefer posting commands like this:
cat file | grep foo | tail | …
# ^^^^^^^^^^^^^^^^^^^ it's easy to copy this
# ^^^^^^^^^^^^^^^^^^^^^ or even this
In other circumstances this would be totally useless use of cat. Some would say it still is, regardless of the circumstances. How about:
< file grep foo | tail | …
# ^^^^^^^^^^^^^^^^^^^ it's easy to copy this
No cat and a convenient form!
Best Answer
It depends on the collation order of your locale:
You can also use the LC_COLLATE variable instead of LANG, and use the POSIX locale instead of the C one.
C collation order is purely alphabetical (ASCII order). Other collation orders (such as English) may consider spaces and special characters such as dots as separators and either handle "words" separately or just ignore these separators (which appears to be the case here).
Note that the non-UTF-8 locale sorts using alphabetic ASCII, too:
After some more digging, it appears that ignoring punctuation is a common feature of Unicode-aware locales such as the
*.UTF-8ones.