glob
glob copied to clipboard
kthompson
Support escaping special characters
Characters like {, }, ,, !,[,] are valid file path names. To match path containing these special symbols, glob pattern need to support escape using \. In this case, \[ should match [ instead of throwing an error.
Hi @kthompson , any update on this? Some of our component right now stucked on this, there are many patterns that contain ( ) as file names but glob don't accept them.
I just released a build that may help. It adds support for all characters that are not special to globbing syntax as part of issue #57
My plan is to add support for escaping characters by using \ which is currently used for a secondary path separator. Since this may break many existing glob expressions this will be part of a 2.0 release.
Hello @yufeih, it has been a while since I have updated this ticket. If you would like to take a look at beta version with this feature you can find it here: https://www.nuget.org/packages/Glob/1.2.0-alpha0029
Hello @kthompson, your library has been very helpful for my projects. Thank you very much!
How would escaping { and } currently be used for delimited literals still work? For example:
` Company\Documents{Ives,Hayes}**.xlsx
In the current(unreleased) implementation, if an escape sequence is detected ie \ then we require one of the following characters:
*?{}[]() or space or , when inside of a literal set to proceed it. Not doing so will result in an exception currently.
As an aside the reason I chose this was because I figured I had the following options:
- Option 1: Stop treating
\as a path delimiter and use it as a strict escape sequence character. If there is not a valid escape sequence throw as an invalid pattern. This breaks existing patterns that use\as a path delimiter but will fail on expression creation. - Option 2: Stop treating
\as a path delimiter and use it as an escape sequence character BUT if there is not a valid escape sequence treat the character as part of a filename(only valid for Linux/OSX?). This breaks existing patterns that use\as a path delimiter but may fail silently - Option 3: Keep treating
\as a path separator but also allow it as an escape sequence, in the event of an invalid escape sequence the\is treated as a path separator. This would be less likely to break existing patterns, but it will still break some(most?) patterns like for example ***.txt previously it would work for any file with .txt but now it would mean anyfile with the literal character * and extension txt
Options 2 and 3 would have the result of having really weird and potentially difficult to spot errors, while Option 1 you would get errors but they would at least be easy to solve.
To get back to your question, if you want to match on a literal set where there is a comma in one of the literals you could do something like a{a\,a,bb} which should match aa,a and abb
Thanks @kthompson. I can understand how this.
I reviewed my previous comment and noticed that I inadvertently dropped the backslash before the curly brace when I was cleaning it up. In this corrected example, I am trying to match a directory that is either "Ives" or "Hayes".
Company\Documents\{Ives,Hayes}**.xlsx
In this scenario, it does not appear that I could have a wildcard immediately after a directory separating backslash.
I was looking around for other ways that this could be done and found in DotNet.Glob - Escaping Special Characters Darrell uses [ and ] instead of backslashes.
@jberd126 I am not completely clear on what you are trying to do but there are a couple issues in your pattern.
** Will only match full directory names if you want a partial match on a directory you need to use *
Second you must use / for path separators in your pattern
Given those two issues you probably want something like:
Company/Documents/{Ives,Hayes}/*.xlsx or Company/Documents/**/{Ives,Hayes}/*.xlsx or Company/Documents/**/{Ives,Hayes}/**/*.xlsx