Automatic escaping characters such as spaces in staged file names

[glormph]

Hi,

Something that trips me up sometimes using NF is when I put quotes around file names, in case they have spaces and other shell-interfering characters, mainly because I'm used to that being good practice. This is is usually fine, but sometimes I find cases where NF escapes these characters, and then the file which is quoted can not be found since it doesn't contain a backslash. So:

• "file name.txt"
• is staged as "file\ name.txt"

I assume that best practice in NF is to not use quotes, as it escapes the relevant characters. Since this sometimes works, I tried to investigate what is going on and got to the following:

• just echo $filename - escapes fine, which I assume represents how it should be
• ${filenames.join(' ') - is not escaped
• $filenames.collect { "$it" }.join(' ') escapes each file, but ONLY if filenames is of length > 1, otherwise it doesn't escape.
• $filenames.collect { "'$it'" }.join(' ') escapes each file (if len>1), but quotes them too -> file not found

Sometimes in "merge" processes I will have a list of undefined length, so I need to deal with this somehow. My somewhat ugly solution is to create a new variable in the process script scope:
filenames_list = instanceof java.util.List ? filenames : [filenames]

I'm not sure if I'm doing this right, and it has been a bit confusing, but I gladly receive any comments or an RTFM link :)

[bentsherman]

In general, you should not quote your filenames because Nextflow will escape any special characters for you.

As for the list issue, that is a long-standing aspect of process inputs that has annoyed many over the years. We intend to fix it with this PR #3706, after which you'll be able to specify arity: true and the input will always be a list if it's a glob. Until then, your instanceof hack is the standard workaround 😄

[glormph]

Thank you for the quick reply! I should learn how to search better, but couldn't find those issues since I looked for quoting and escaping instead of glob/list/path outputs.