rsync all pdfs except in certain directories?

I'm trying hard to understand the rsync filter system, and it's completely baffling me.

I have the following "test" directory structure to try to make sense of it. With no filter options here are all my files:

rsync -amv --dry-run /source /target

building file list ... done
source/
source/1.pdf
source/2.pdf
source/exclude_rules.txt
source/filter_rules.txt
source/excludedir/
source/excludedir/2.jpg
source/excludedir/4.pdf
source/subdir/
source/subdir/1.jpg
source/subdir/1.txt
source/subdir/3.pdf
source/subdir/subdir2/
source/subdir/subdir2/6.jpg
source/subdir/subdir2/6.pdf

I just want to sync all *.pdf files except in certain directories, namely any directory that has *exclude* in it.

I'm using a file with the filter rules in it with the following command:

rsync -amv --dry-run --filter='merge /filter_rules' /source /target

The filter_rules look like variations on the following but I can't get them to produce the results I'm after:

-/ *exclude*/
+/ *.pdf
-/ *

The closest I've come is with the simple exclude:

-/ *exclude*/

Which yields:

building file list ... done
source/
source/1.pdf
source/2.pdf
source/exclude_rules.txt
source/filter_rules.txt
source/subdir/
source/subdir/1.jpg
source/subdir/1.txt
source/subdir/3.pdf
source/subdir/subdir2/
source/subdir/subdir2/6.jpg
source/subdir/subdir2/6.pdf

How do I filter the rest to just get *.pdf ?

For posterity, I did finally get this to work, and here are the instructions I wish I had had:

• rsync starts the filter process with a full list of files
• the filter rules are handled IN ORDER (took me a while to get this)
• You may have all the right rules, but not the right order, so if you're using external exclude or include files, they may need to be re-ordered with a filter file which allows you to mix and match include/exclude rules, or listed on the cli itself
• for each file, The FIRST FILTER RULE THAT MATCHES puts the file into one of 2 buckets, include or exclude.
• Rules after the first matching rule are not applied!
• Each rule acts only on the files that made it "past" the previous rules not being matched
• Files that don't match any rules are INCLUDED
• The last rule is the most important and unintuitive, and it means exclude everything that wasn't specifically included UP TO THAT POINT.

So here's what ended up working:

-/ *exclude*/
+/ */
+/ *.pdf
-/ *

Originally I had those rules in separate include-from and exclude-from files, and that wouldn't allow for the proper order.

I still use --exclude-from in my rsync, but this link was remarkably helpful when I attempted to get filtering to work.

https://stackoverflow.com/questions/35364075/using-rsync-filter-to-include-exclude-files

edit - the OP nailed this in his own answer, but as requested... helpful bit from that link

Explanations:

(only rewording the manual in the end but as you said the manual is a bit >cryptic)

Rules are read from top to bottom each time a file must be transferred by >rsync. But in your case /mnt/data/i-want-to-rsyncthisdirectory/ is not >backed up because you exclude /mnt and this short-circuits your include >rules. So the solution is to include each folder and subfolder until the >folder you want to back up and then to exclude what you do not want to back >up subfolder by subfolder.

Note the * at the end of each subfolder exclusion. It will prevent rsync to >back up the files and folder located in these subfolders which is what you >want I think. Simpler solution: (edit 2)

You can even simplify this with the *** pattern that was added in version >2.6.7:

• /mnt/
• /mnt/data/
• /mnt/data/i-want-to-rsyncthisdirectory/***

• /mnt/**

This operator allows you to use the ** wildcard for exclusion and >consequently to have only one exclude line.

I also discovered that you can understand which filter rules exclude/include >each file or folder thanks to the following rsync arguments:

--verbose --verbose

Combined with the --dry-run argument you should be able to debug you problem >:)