Complex file search via command line

pico77 · 10-12-2017, 11:05 AM

Dear Users

I have pretty basic usage of the common 'find' and 'locate' commands which I prefer over other tools for searching files/folders.

I know there are indexing software like recoll or beagle, but they put out too many results. I name files/dirs in a very structured way: the typical file/directory name in my system is

2017_10_12_abc_def_ghi

it starts with an important date (usually the one I started/finished to work on the file), "abc", "def", "ghi" are descriptors/keywords that tell me about the content according to some hierarchy of mine. Let's say that any string separated by underscores is a Keyword/descriptor in my naming scheme.

The problem is that I'd like to construct a script (maybe based on find/locate) that does the following: in a given path it looks recursively to find all files/dirs whose names descriptors contain a list of possible strings

For example the following

Code:

command_name   path_name    "ab"  "hi"

should look recursively into 'path_name' and search for all files/dirs that have keywords containing at least one of the two srings (that is "ab" or "hi"). So for instance such a command should give

2017_10_12_abc_def_ghi

as a result, but not

2017_10_12_cba

I tried to construct such a thing with find/locate commands, but I might be too basic to come up which something working. Wha it is most difficult to me is that the list of possible keaywords over which to look should be of variable length.

Does anybody have some idea if this is possible to do?

Regards
Pierre

nodir · 10-12-2017, 01:43 PM

second code block here:
http://mywiki.wooledge.org/UsingFind...based_on_names

Quote:

Wha it is most difficult to me is that the list of possible keaywords over which to look should be of variable length.

I don't understand that.

dugan · 10-12-2017, 01:50 PM

https://github.com/junegunn/fzf

tofino_surfer · 10-12-2017, 02:57 PM

Quote:

Does anybody have some idea if this is possible to do?

With perl it is quite simple to do. You need an actual text processing scripting language such as perl for this. Perl has its own find function which will find all filenames recursively from a starting point and call a function that you specify for each file name. You would use this function as a filter. You would pass a list of strings to match as an argument to the script which will be used by this filter function. If the current filename matched one of these strings you would add it to an output list. When the find function was finished scanning you would output this list.

The script you write will accept a directory name as the first argument and a variable number of strings to be used as filters. The directory will be the starting point for the find function while the strings will all be added to a global variable list which would be referenced by the filter function.

http://perldoc.perl.org/File/Find.html

You will have to learn some perl to do this but with an actual scripting language this is quite straightforward.

ondoho · 10-13-2017, 01:27 PM

and why is

Code:

find /path -name '*ab*hi*'

not sufficient?

pico77 · 10-13-2017, 01:53 PM

Quote:

Originally Posted by dugan

https://github.com/junegunn/fzf

this is very interesting. I Installed it via github, but I need to understand how to use it.

thanks for this.

P

pico77 · 10-13-2017, 01:56 PM

Quote:

Originally Posted by ondoho

and why is

Code:

find /path -name '*ab*hi*'

not sufficient?

yes, in principle yes, but the point is that this would find relevant items only if the order of the keaywords is correct. In other words

Code:

find /path -name '*ab*hi*'

will only find items where 'ab' comes before 'hi'.

Thanks anyway
P

ondoho · 10-13-2017, 02:16 PM

ok, maybe like this:

Code:

find /path -name '*ab*' -a -name '*hi*'

tofino_surfer · 10-17-2017, 10:43 PM

The OP was looking for some sort of script to avoid typing a complex command hundreds of times. The following bash script builds a variable length find command adding -o -name for each parameter past 2.

Code:

#!/bin/bash

FIND="find $1 -name '*$2*'"

for item in "${@:3}" ; do
  FIND="$FIND -o -name '*$item*'"
done

#echo $FIND
eval $FIND

With echo $FIND un-commented for debugging.

Code:

$ ./search.sh /home/pico ab hi mr ty
find /home/pico -name '*ab*' -o -name '*hi*' -o -name '*mr*' -o -name '*ty*'

This will greatly cut down on the keystrokes required.

syg00 · 10-17-2017, 10:57 PM

I like locate with regex - much kinder on the machine. Especially my poor old spinning disk.

Code:

locate --regex "$PWD.*(ab|hi)"

pan64 · 10-18-2017, 04:00 AM

Yes, I think you won't find better than that locate command with --regex. But you can also use ls -lR <dir> | grep pattern instead of find, it will be much faster.

ondoho · 10-18-2017, 11:35 AM

^ doesn't locate find only files that have been indexed in its database?

syg00 · 10-18-2017, 06:27 PM

Sure.
Some some distros run updatedb daily, so that might suffice for most folks. Me, I run it when I feel the need - usually less than once a week.