The command you've posted in the question
find . -type d -execdir bash -c "mkdir -p subdirectory_a && mkdir -p subdirectory_b && mv *_a.tif subdirectory_a 2>/dev/null && mv *_b.tif subdirectory_b 2>/dev/null" \;
behaves correctly ( that is , it does what it is supposed to do ). But what it spells out is not what you want.
The -execdir argument operates such that "...the specified command is run from the
subdirectory containing the matched file," (source, find(1) )
That is to say, when your current working directly is root and the command locates subdirectory1 it will not descend there as you expect. The subdirectory1 which matches -type d was found when find was inside root, hence it will stay there and execute the mkdir command specified from there. Naturally, there is no files inside root, hence nothing will be moved around, but also note that mkdir -p won't produce an error, as specified by the mkdir(1) : "no error if existing".
So how do we make this match what you want? Let's assume the same structure you've posted
$ tree /tmp/root
/tmp/root
├── subdir1
│ ├── file_a1_a.tif
│ ├── file_a1_b.tif
│ ├── file_a2_a.tif
│ └── file_a2_b.tif
└── subdir2
├── file_b1_a.tif
├── file_b1_b.tif
├── file_b2_a.tif
└── file_b2_b.tif
I'd suggest making it a three-step process. There is no need to make a one single command do all the work. Focus on the desired result rather than "code golf". So make the directories first
$ find /tmp/root -maxdepth 1 -mindepth 1 -exec bash -c 'mkdir ${1}/subdirectory_{a,b}' sh {} \;
$ tree
.
├── subdir1
│ ├── file_a1_a.tif
│ ├── file_a1_b.tif
│ ├── file_a2_a.tif
│ ├── file_a2_b.tif
│ ├── subdirectory_a
│ └── subdirectory_b
└── subdir2
├── file_b1_a.tif
├── file_b1_b.tif
├── file_b2_a.tif
├── file_b2_b.tif
├── subdirectory_a
└── subdirectory_b
6 directories, 8 files
And then find regular files at specific level in directory tree and move them. Now we can use -execdir to achieve exactly what we want
$ find /tmp/root -maxdepth 2 -mindepth 2 -type f -execdir bash -c 'case ${1} in *a.tif) echo cp ${1} ./subdirectory_a ;; *b.tif) echo cp ${1} subdirectory_b ;; esac' sh {} \;
cp ./file_b2_b.tif subdirectory_b
cp ./file_b2_a.tif ./subdirectory_a
cp ./file_b1_a.tif ./subdirectory_a
cp ./file_b1_b.tif subdirectory_b
cp ./file_a1_b.tif subdirectory_b
cp ./file_a1_a.tif ./subdirectory_a
cp ./file_a2_b.tif subdirectory_b
cp ./file_a2_a.tif ./subdirectory_a
Note that in this case, I use echo before moving to ensure everything behaves properly , and no actual moving is done. So if we remove echo, this will behave correctly:
$ find /tmp/root -maxdepth 2 -mindepth 2 -type f -execdir bash -c 'case ${1} in *a.tif) cp ${1} ./subdirectory_a ;; *b.tif) cp ${1} subdirectory_b ;; esac' sh {} \;
$ tree /tmp/root
/tmp/root
├── subdir1
│ ├── file_a1_a.tif
│ ├── file_a1_b.tif
│ ├── file_a2_a.tif
│ ├── file_a2_b.tif
│ ├── subdirectory_a
│ │ ├── file_a1_a.tif
│ │ └── file_a2_a.tif
│ └── subdirectory_b
│ ├── file_a1_b.tif
│ └── file_a2_b.tif
└── subdir2
├── file_b1_a.tif
├── file_b1_b.tif
├── file_b2_a.tif
├── file_b2_b.tif
├── subdirectory_a
│ ├── file_b1_a.tif
│ └── file_b2_a.tif
└── subdirectory_b
├── file_b1_b.tif
└── file_b2_b.tif
6 directories, 16 files
Note of course that this uses copying to avoid data loss. We can further clean up and in the end - we get the desired directory tree structure.
$ find /tmp/root -maxdepth 2 -mindepth 2 -type f -delete
$ tree
.
├── subdir1
│ ├── subdirectory_a
│ │ ├── file_a1_a.tif
│ │ └── file_a2_a.tif
│ └── subdirectory_b
│ ├── file_a1_b.tif
│ └── file_a2_b.tif
└── subdir2
├── subdirectory_a
│ ├── file_b1_a.tif
│ └── file_b2_a.tif
└── subdirectory_b
├── file_b1_b.tif
└── file_b2_b.tif
6 directories, 8 files
In conclusion what we've done can be expressed as one single script
find /tmp/root -maxdepth 1 -mindepth 1 -exec bash -c 'mkdir ${1}/subdirectory_{a,b}' sh {} \; &&
find /tmp/root -maxdepth 2 -mindepth 2 -type f -execdir bash -c 'case ${1} in *a.tif) cp ${1} ./subdirectory_a ;; *b.tif) cp ${1} subdirectory_b ;; esac' sh {} \; &&
find /tmp/root -maxdepth 2 -mindepth 2 -type f -delete
Of course, steps 2 and 3 can be condensed into one if we use mv instead of cp, but I would rather err on the side of caution than to risk data loss.
