Crass was designed primarily for short read data like the kind produced from the Illumina sequencing platform. However it has been tested on Sanger and Ion Torrent data as well. The current constraints are that reads shorter than 76bp in length will not be used and reads longer than 2000bp cannot be used
No, currently Crass will only work on the raw sequencing reads. If you have access to the data you could always run Crass on the reads that went into making that genome of yours. There are three programs that will identify CRISPRs in genome sequences: PILER-CR, CRT & CRISPRFinder; you might want to check them out. In the future Crass will support full length genome sequences, however due to assumptions made in the algorithm it is not currently possible.
Well that depends on how much data you have and how diverse that data is. The algorithm used in Crass scales linearly with the number of reads processed but exponentially with the number of CRISPR patterns in the dataset. With a dataset with 100M reads and a medium level of CRISPR diversity, Crass will run overnight and be read in the morning.
Best read the tutorial
Currently no, but it is an area of improvement for the algorithm
For starters, make sure that you are using the latest release of Crass (the most recent version number is found all over this website). The next thing to do is to make sure that it is reproducable on the dataset that your using, so just run Crass again to find out (and make sure that you turn the logging up to the maximum level you can). Assuming that you are on the latest version and you can reproduce the bug you should log the bug here. It's important that you give as much information about the error including pasting any text that was output to the terminal. You should hang on to the log file cause I may ask for it when I'm debugging