Re: [Gmod-ajax] Jbrowse JSON files that are not JSON.

[Mitch S. <mit...@be...>]

Martin A. Hansen wrote:
> I am wondering why Jbrowse files like refSeqs.js and trackInfo.js are 
> not in JSON format, but rather have a variable name, an equal sign, 
> and then, on the following lines, a block of JSON code:
>
> refSeqs =
> [
>    {
>       "length" : 50001,
>       "name" : "ctgA",
>       "seqDir" : "data/seq/ctgA",
>       "seqChunkSize" : 20000,
>       "end" : 50001,
>       "start" : 0
>    }
> ]
>
> Why not encode the entire thing in JSON?

It's because of the caching behavior of XMLHTTPRequest in (at least some 
versions of) IE.  For files that the browser knows about directly (files 
that show up in <script> elements in the HTML), the browser will check 
if they've changed when the user refreshes the page.  But in IE, files 
that were fetched using XMLHTTPRequest always come from the cache 
(unless you add a query parameter that changes with each request so that 
the file never comes from the cache).

I wanted the client fetch the file if (and only if) it had changed, so I 
figured I'd add an assignment line like the one above and put the file 
in a <script> element rather than getting it via XHR.

Also, getting the refseq and track list information using a <script> 
element means that the data will already have been fetched when the 
embedder calls the Browser constructor.  On the other hand, you could 
argue that that constructor should just be getting URLs and doing its 
own fetching, if (say) we wanted to refresh the track list without 
refreshing the page.

As you've noticed, the track data and name tree files are just straight 
JSON, and those are fetched with XHR, so I guess there's a chance of IE 
getting stale data if one of those files has changed.  Googling just 
now, it looks like there may be some nicer workarounds for this IE bug:

http://en.wikibooks.org/wiki/JavaScript/XMLHttpRequest#Caching

so changing this in JBrowse is probably something we should do at some 
point.

Mitch