Re: Request: Test suite for EFS.

[James Antill <james-gnome-components and org>]

Miguel de Icaza <miguel@gnu.org> writes:

 I'm going to jump in here, although I mostly just lurk (on the
components list) and most of my work doesn't involve GNOME/XML.

> But how do you address standard naming schemes with the XML file
> format, for things like "Summary/Author" in a way that is consistent
> across applications.  How can we standarize this?

 I'm not sure I understand what you mean. Surely just mandating...

<summary><author> blah </author></summary>

...the same way you would mandate a summary/author "file" in efs.

> How can I quickly scan information?  How can I pull an image from
> another XML file without having to load the entire XML file into
> memory?

 Presuming no indexes, an API like (sorry if the naming scheme is
wrong)...

GnomeXmlFastScan *scan = gnome_xml_fast_scan_create( ... );

gnome_xml_fast_scan_reg(scan, "<summary><author>", ... );
gnome_xml_fast_scan_reg(scan, "<summary><comment>", ... );

gnome_xml_fast_scan_file(scan, "filename", ... );

 At which point _only_ <summary><author> and <summary><comment>
contents would be guaranteed to be valid. Providing the ability to
"upgrade" from a fast scan'd file to a fully parsed file might be
useful but isn't necassery.

 The _reg command might want to be "summary/author" (to fit in with
the menus etc.) or need to have the whole 
"<summary><author></author></summary>" for whatever reason. The idea
should be obvious from above though, whatever the implementation.

 As for the write performance, I would be _very_ supprised if just
writting the whole file out and calling rename() wasn't fast enough
(and it's also going to be correct if multiple apps write to one file
or you use it over NFS etc.).

> (Consider embedding an image which is inside a document into a
> spreadsheet without activating the document component)

 This just requires that everyone use the same SS system no ? --
Ie. the image can be extracted either way as long as the container
(sp?) knowns how to read the components files.
 Indeed this will be much easier for perl/python/etc. because they can
use the XML primitives to load the parts fo the document they want.

 Ps. Recent to components list seperatly because I didn't know you had
to mail from the address you're subscribed from.

-- 
James Antill -- james@and.org
I am always an optimist, but frankly there is no hope.
   -Hosni Mubarek