A lot of my projects include reading layer names from DWG-files and check if they are consistent with a layer standard or if they are empty. As I'm both interested in the data features (for other tests) and the schema I need to retrieve both from the DWG files. There are various ways of doing this and the different techniques varying in speed pretty dramatically.
For my test purpose I created a DWG-file (Release2013) with 500 point features, each on an unique layer (so the DWG-file has 501 layers with layer 0). Using FME 2017.1.2.0. Win 64.
- Reading the Schema with a FeatureReader transformer, Autodesk AutoCAD DWG/DXF-reader, just reading the Schema Features. Group Entities By "Layer name": 5 minutes 21.6 seconds
- Reading the Schema with a FeatureReader transformer, Autodesk AutoCAD DWG/DXF-reader, just reading the Schema Features. Group Entities By "Attribute Schema": 35.5 seconds (this options is useless for me as the Schema reader just read layers that holds entities and I need to check which layers that are empty)
- Reading the Schema with a FeatureReader transformer, Schema (Any Format)-reader, choosing Autodesk AutoCAD DWG/DXF as input format: 3.1. seconds
- Reading the Schema with a Schema (Any Format)-reader, choosing Autodesk AutoCAD DWG/DXF as input format: 3.2 seconds (one interesting thing here is that the reader says Successfully opened the 'Release2018' AutoCAD file even if the file is a Release2013 file)
(there were no major time differences when choosing the RealDWG-reader/format instead)
This means that the most efficient way of reading the Schema is by using the Schema (Any Format)-reader stand alone or inside a FeatureReader transformer. This means that I need two Readers/FeatureReaders in my workspace as I'm also interested in the data features. This seams a bit unnecessarily as the FeatureReader can read the Schema while I'm reading the data.
Is there any chance of tweaking the Schema reading inside the FeatureReader transformer while reading the data with the Autodesk AutoCAD DWG/DXF-reader so it gets a big faster?
(the test dataset has 500 layers but the real files I verify right now has over 1700 layers...)
