What expressions would match the pattern (^[0-9]..[a-zA-Z ]+$) in grep command? Linux bash

Question

Im trying to understand what expressions exactly the regular expression (^[0-9]..[a-zA-Z ]+$) detects in grep command (linux terminal)

I know that if I'd write the following command:

grep ^[0-9]..[a-zA-Z] filename.txt

I will detect any line that contains expressions such as 92afg But Im not sure what the +$ means and what kind of expressions will I be able to detect with the command

grep ^[0-9]..[a-zA-Z]+$ filename.txt

I tried to open a new text file and just type expressions that I thought would be detected, but none of them matched, so I'd appreciate explanation for this.

Answer 1

Let's break it down. First of all, note that this RegExp uses the "Extended regular expression" syntax (ERE) - the + is a metacharacter that doesn't work in the "Basic regular expression" syntax that grep uses by default (meaning it would match itself and require a literal + at that position), so if you want to use that RegEx with grep, you will need to pass the -E option.

• The ^ is an anchor that ties this position of the regex to the start of the line.
• The [0-9] is a character list and will match any single⁽¹⁾ character that falls into the sort range between 0 and 9. What exactly that comprises depends on the "collation order", determined among others by the environment variable LC_COLLATE.
• The . matches any single character, so two .. means "any two characters".
• The [a-zA-Z] again is a character list and will match characters⁽¹⁾ that fall between a and z and in addition those that fall between A and Z. Again, what that means depends on the collation order!
• The + means "one or more of the previous"
• The $ is an anchor that ties this position of the regex to the end of the line.

So, your RegEx is intended to⁽¹⁾ match any lines that

• start with any digit
• followed by any two characters
• and only contain letters (but at least one) up to the end of the line.

⁽¹⁾for what it might actually do, see below

Some notes

1. In your example, you use the regular expression unquoted. That means any characters are open to interpretation by the shell before they are passed to the grep command. If your pattern contains $ or globbing characters (*, ? and [...] character lists!), the shell may try to perform variable expansion (thereby replacing parts of your RegEx) or expand globbing patterns into possibly multiple filenames, so that in the end you would have more arguments on the command-line that you originally intended. Other characters that are special to the shell (>, #, ; and the like) might lead to even more unexpected behavior. You should use

   grep -E '^[0-9]..[a-zA-Z]+$' filename.txt
   

   instead. Note that you can get rid of the opening and closing anchors by using the -x flag to enforce "whole-line" matching:

   grep -x -E '[0-9]..[a-zA-Z]+' filename.txt
   

2. Character lists containing ranges (such as a-z) are dangerous because they might not give you what you think. Naively one might expect them to match all characters that lie between the start and end character on the ASCII table, but that is only true for the C locale. In other locales (and in particular in the usually set system locales such as en_US.UTF-8) the collation order is something like aAbB ... zZ so a-z would also match most upper-case letters. Also, the match is actually not on the level of single characters but "collation elements" which means in some locales, even combinations of several letters may match (e.g. dzs in Hungarian)! See this answer (or, in general, most answers by @Stéphane Chazelas about pattern matching) for more insight. If you want to ensure that your ranges work, set the collation order at least for the given command via

   LC_COLLATE="C" grep -E ' ... ' filename.txt
   

Answer 2

+ stands for "one or more repetitions of the previous", $ is "end of line". Note the difference versus * with means "zero or more repetitions".

So it basically means: Any line starting with a digit, followed by two characters of any kind and subsequently one or more (possibly capital) letters¹ until the end of the line.

(¹ be careful, some locales might not only have the 26 letters you'd expect in A-Z or a-z, e.g. è or ŷ depending on language)

For a good guide regarding regexes, I strongly suggest grymoire's beautiful website, which I heartly also recommend for e.g. sed and awk.

---

Why doesn't it match?

+ is part of the extended regular expressions (and otherwise is interpreted as a literal +-sign).

So for using + as "one or more repetitions", use the -E-flag in grep and also quote the regex to avoid any issues with shell special characters:

grep -E '^[0-9]..[a-zA-Z]+$' filename.txt