Hi @bobo, actually a File Geodatabase dataset is a set of files contained by a folder whose name ends with ".gdb". You can confiture the PATH reader to read only "*.gdb" by setting its parameters, as in:

Got it worked, but still, there's a problem with featurereader, I use directory and file pathnames to read 17 gdbs, and then connected output to featurereader which also I set to read same 17 gdbs. It seems that feature will output all data 17 times through generic port, but I only need all data to be read one time. I think maybe I understand one of transformers incorrectly.

Since the PATH reader outputs features each of which has an attribute called "path_windows" that stores a *.gdb folder path (i.e. a File Geodatabase dataset location), you can set it to the Dataset parameter in the FeatureReader to read the datasets one by one.

In the case where you can set all the *.gdb paths to the Dataset parameter statically, you don't need to use the PATH reader. Just use a regular File Geodatabase reader simply.

Problem solved, thank you! Liked I said, I put a Tester between Directory and File Pathname and FeatureReader, and I created a published parameter to control workflow, Tester passed to continue. Regular FGDB reader will have to read every feature from gdb and each gdb has over 1 million feature. It will take too much time if I save workbench in fme extension. Directory and File Pathname and FeatureReader give me control of which gdb to be read. BTW, is there a easier way to do so?