Yes that is definitely much too long. If you use validation 'limited' and output to EXI is the performance roughly similar? (Probably, but worth checking)
The other is making sure the JVM has enough memory. If the JVM is borderline at the memory limit it will spend a large fraction of its time garbage collecting. I would be happy to take a look at your schema for performance tuning suggestions confidentially and outside of this forum. For example, there are often missing discriminators that cut out huge amounts of backtracking. (Especially in textual data formats with multiple kinds of records.) That's a common omission where the schema works and is technically correct, but because the automatic optimization of this is quite difficult (sometimes not feasible at all), Daffodil executes these much more slowly than it otherwise might. On Fri, Apr 14, 2023 at 1:37 PM Roger L Costello <coste...@mitre.org> wrote: > The input file is 375 MB > The XML file that DFDL parsing generated is 4.67 GB > > Time required for Daffodil to parse the input and generate the XML file is > 16 minutes, 24 seconds. > > Ugh! > > That is too long. >
