Re: Error occurring in dods.dap.Parser

On Jan 26, 2005, at 9:30 PM, Tennessee Leeuwenburg wrote:

> Unfortunately that seems to break thredds's parsing of  
> catalogConfig.xml.

Hmmm. If you look in Java-DODS/dods/dap/parser/DDSJava.jj you can see  
the characters that the scanner gobbles up. I'm not really sure how the  
THREDDS code works.

If you have the parser generator, you could try adding '=' and ',' to  
the set of allowed chars. You'd need to be careful though because = and  
',' are used by the CE syntax. So you can have them be part of a  
dataset name, but if they are a variable name the CE parser will become  
very confused.

A more sweeping solution (besides switching to XML, which is in the  
works) would be to modify the parser so that it uses a double quote (")  
to quote chars so any char could be part of a word (you'd do the usual  
thing about providing for quotes in the set of chars itself). This  
would be a more drastic change to the parser but would also dovetail  
with the current parser (I think) and not require a switch to XML  
(which solves the problem because it provides for robust quoting).

James


> If I convert
> <dataset name="Single-level Salinity Data"  
> urlPath="stream=ocea,param=200,DATE=20000101,TIME=0000,STEP=0,EXPVER=0o 
> c7,CLASS=RD,TYPE=AN,NUMBER=0,METHOD=0,LEVELIST=7.000,LEVTYPE=DP,REPRES= 
> LL,DOMAIN=G,salinity.nc"/>
>
> to
>
> <dataset name="Single-level Salinity Data"  
> urlPath="stream%3Docea,param=200,DATE=20000101,TIME=0000,STEP=0,EXPVER= 
> 0oc7,CLASS=RD,TYPE=AN,NUMBER=0,METHOD=0,LEVELIST=7.000,LEVTYPE=DP,REPRE 
> S=LL,DOMAIN=G,salinity.nc"/>
>
> that is to change the first "=" to %3D, the hex code for "=", it seems  
> to break the lookup. What I get is the error
>
> Error {
>    code = 1;
>    message = "Dataset not found in catalog; urlPath=  
> <stream=ocea,param=200,DATE=20000101,TIME=0000,STEP=0,EXPVER=0oc7,CLASS 
> =RD,TYPE=AN,NUMBER=0,METHOD=0,LEVELIST=7.000,LEVTYPE=DP,REPRES=LL,DOMAI 
> N=G,salinity.nc>";
> };
>
>
> It would seem that thredds is converting the %3D to "=" in its head,  
> then failing on the string comparison. Just my guess though.
>
> Cheers,
> -T
>
> > > T,
> > >
> > > Well. The problem is in your Dataset name:
> > >
> > > > }   
> > > > stream=ocea,param=200,DATE=20000101,TIME=0000,STEP=0,EXPVER=0oc7,CLA 
> > > > SS  
> > > > =RD,TYPE=AN,NUMBER=0,METHOD=0,LEVELIST=7.000,LEVTYPE=DP,REPRES=LL,DO 
> > > > MA IN=G,salinity.nc;
> > >
> > >
> > > The parser is barfing on the first "=" character (Line 58, column 9)
> > >
> > > James? Is this in any way a legal name? If so, how does he need to   
> > > "escape" the "=" (and possibly the ",") characters?
> >
> >
> > You could escape the '=' and ',' using the HTTP %<hex digit><hex  
> > digit>  notation, I think. When the Java DAP 2.0 parser was written  
> > we had not  encountered file names that broke the 'C identifier'  
> > mold. But that was  just around the corner... Anyway, the C++ code  
> > lets just about anything  be a file name. I'd try the escaping scheme  
> > and see how it works.
> >
> > James
> >
> > PS. I just checked the DDS parser and it will grok the percent sign,  
> > so  you should be OK using that to signal the start of the escape  
> > sequence.
--
James Gallagher                jgallagher at opendap.org
OPeNDAP, Inc                   406.783.8663