0

I have a Bash function f, which takes a string value s as one of its arguments.

I need to abort execution if s is not an integer between [1000-9999].

I initially thought I would try to cast the string to an int and perform arithmetic range checks on the result, aborting on errors, but all other parameters are checked using RegEx so I might as well be consistent. I have to, actually.

Bash version: 4+

I could not come up with a better-looking pattern than this:

[[ "$s" =~ [[:\<:]][[:digit:]]{4}[[:\>:]] ]]

Can you help me improve the readability of the pattern-check construct above?

What I particularly dislike in the pattern above is:

  • having to backslash escape the word boundary character < as [[:\<:]]
  • the sheer length of it, and the insane number of square brackets

One idea I had was to use full-line boundary checks, although that's a hack? (i.e. using ^...$ rather than \b...\b)

Optionally :: Is there a fairly well-established pattern for argument validation in Bash?

Improve this question
12
  • If "consistent" entails "unreadable", ditch consistency. Commented Mar 10, 2013 at 7:00
  • @n.m. Well.. is it impossible to make the pattern above readable? "i.e. match 4 integers and 4 integers alone" Commented Mar 10, 2013 at 7:02
  • 1
    If I had to use regexes I would ditch posixized character classes and use plain old ^[0-9]{4}$. Don't see ^$ as a hack btw. Commented Mar 10, 2013 at 8:26
  • @n.m.: I think your argument is good and you are convincing me.. I am going to upvote your comment now and accept your answer if you care to enter it as such. (I was indeed going for POSIX "parlance" but thinking about it, I don't really need it) Commented Mar 10, 2013 at 10:54
  • Bash uses ERE, which doesn't support word boundaries (unless you're using a tool like GNU grep) Commented Mar 10, 2013 at 11:58

3 Answers 3

Reset to default
2

Also, I would try:

[[ $s =~ ^[1-9][0-9]{3}$ ]]

Otherwise 0999 would pass. But an other aspect might be that +1234 would probably be valid input, and so would perhaps be 1234.0, so if you need to be complete you may need to expand your regex, or perform value checking after all..

Improve this answer
Sign up to request clarification or add additional context in comments.

2 Comments

One thing I really appreciate is this kind of preciseness. I even admire it. You are very right, for this to be correct it cannot be [0-9]{4} but it should be more something like you say. Very good attention to detail.. upvoting!
Yeah, this seems much simpler than the others.. accepting! :)
2

bash can do integer comparisons:

shopt -s extglob # Needed for the extended pattern +(...)
f () {
    if [[ $1 == +([[:digit:]]) ]] &&
       (( $1 >= 1000 && $1 < 10000 )); then
    ...
    fi
}
Improve this answer

9 Comments

I like the +(pattern) notation, as I specify it's Bash anyway. Upvoting because of that. I also like the short-circuited &&: very correct. I dislike the BIG if clause encompassing the entire function, although some people like to exit from just one location in a function I think returning at the top after argument validation is simple and readable. BTW: shouldn't you be quoting the var expansion?
You can invert the sense of the test ( if [[ $1 != ... ]] || (( $1 <1000 || $1 > 10000 )); then), and just handle the error condition in the body of the if statement. Parameter expansions don't undergo word splitting inside [[ ... ]], so you don't need to quote them.
wow, did not know that expansions don't undergo word splitting inside [[ ... ]]. So this is robust against strings like "a1000", "a 1000", etc.? I say so because you don't use word boundaries there..
All && and ||'s short-circuit. Within [[ ]] and (()) as well. Just not for expansions within (()).
@chepner: is there anything, in your opinion, that your more elaborate code does that Scrutinizer's solution does not do, though?
|
1

See http://mywiki.wooledge.org/BashFAQ/054

You could do this using only a pattern match but I'd split it. I would say something like this is close to the "usual" method:

f() {
    if [[
        $1 == +([[:digit:]]) &&
        "$1 >= 1000 && $1 < 10000" -ne 0
        ]]
    then
        ...
    fi
}

In Bash/ksh test expression compound commands, the arguments to numerical comparison operators are evaluated as arithmetic expressions. The double quotes allow for whitespace within the expression, otherwise the parser can't distinguish some arithmetic operators from test expression operators. Testing for whether the expression on the left is 0 is basically equivalent to (()). This is identical to @chepner's answer, except using only one command. I'd expect this performs slightly better and I'm used to the grammar so it's pretty clear to me. Some prefer the other way. You should only use the [[ arithmetic operators over (( when combined with some other test.

You have to first validate before using any unpredictable input as an arithmetic expression. Then you can check for whether it's within a certain range. There are other ways to do this, but this is how I would do it personally. It's clear and works for all permutations of this problem.

I should also add that if your input might contain octal or hex literals, some further processing is needed. Usually running the input through printf %d works well in that case.

Improve this answer

4 Comments

I don't see this construct much: "$1 >= 1000 && $1 < 10000" -ne 0; can you elaborate on your rationale for enclosing in double quotes the entire check and the use of -ne? You can edit the answer. Evidently I may profit from learning a new "thinking pattern" here.. for now I am just confused by it :(
Hopefully that explains it. Another fun fact: mksh and pdksh uses arithmetic expressions in this way even for plain [ tests, and its -a and -o operators short-circuit in that case. :)
Ah-a! I learned something new. Upvoting in gratitude. Still not sure this is any more "correct" than Scrutinizer's solution though?
It's correct in general for this kind of input. That way is perfectly fine for this specific situation. It's whatever you prefer. I like this way because I immediately recognize it as checking for a valid number.

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Start asking to get answers

Find the answer to your question by asking.

Ask question

Explore related questions

See similar questions with these tags.