NOTICE: This version of the NSF Unidata web site (archive.unidata.ucar.edu) is no longer being updated.
Current content can be found at unidata.ucar.edu.
To learn about what's going on, see About the Archive Site.
On 3/16/2012 12:02 PM, Dennis Heimbigner wrote:
1. It reads DDS, DAS outputs and then attempts to read the contents of all string data variables in DDS. Since some strings may be very long, it degrades the performance. I don’t know why it tries to access string variables when it opens data via OPeNDAP. Reading DDS and DAS seems sufficient.
Its because netcdf data model has char datatype but dap2 has String, and theres no way to tell the length of a string without reading.
prefetching strings was a compromise on fixing this problem. it can go away with DAP4.
The client side code attempts to do some prefetching of small variables.The hidden assumption is that string values are not very big, and indeed this is usually the case. I am curious what kind of dataset you have that violatesthis assumption.2. If a string size is more than 32767, it throws an error. I don’t understand why it put such restriction yet. The line 151 and 152 of [1] checks Short.MAX_VALUE, which is 2^15 - 1 = 32767 according to [2]. This restriction makes it fail to access otherwise valid OPeNDAP data.
dennis can you check this? is 32767 limit part of dap2 spec?
3. It forms a very long URL string if data file has many variables with long variable names. If the requested URL is too long for server to handle, it returns an error “opendap.dap.DAP2.Exception: Method failed:HTTP/1.1 400 Bad Request on URL=http...<all variables will be listed here>”. I don’t understand why it tries to append all variables to get DDS output. Here's the part of code that constructs a very long URL:
Joe, can you send a URL with this problem?
This occurs primarily because of prefetching of "small" variables.Later, it may need to get the whole DDS (for display in toolsUI, for example) so it forms a URL requesting all variables except those already pre-fetched.=Dennis Heimbigner Unidata. _______________________________________________ netcdf-java mailing list netcdf-java@xxxxxxxxxxxxxxxxFor list information or to unsubscribe, visit: http://www.unidata.ucar.edu/mailing_lists/
netcdf-java
archives: