Re: [xml] need info on using libxml2 - build, merge, template xml files

[Rush Manbert <rush manbert com>]

Senthil Nathan wrote:
> Hi Rush,
> Gr8. Thanks for the info.
>
> And reg. the 4th point, I need to insert the values in the proper places of
> the DOM tree.
> You got it rightly. So how do I get that done??
>
> Senthil
>
> On 8/17/07, Rush Manbert <rush manbert com> wrote:
>
> > Senthil Nathan wrote:
> >
> > > Hi,
> > > I'm new to use libxml2. I need general clarifications of using libxml2
> >
> > for
> >
> > > my application.
> > >
> > > Say, I have a XML content as,
> > >
> > > <hostName>
> > >      <defaultValue>value</defaultValue>
> > >      <description>
> > >             <shortDescr>Hostname descr.</shortDescr>
> > >             <longDescr>long descr of Hostname</longDescr>
> > >      </description>
> > >
> > >      <dataType>string</dataType>
> > >      <operations>set,delete</operations>
> > >
> > > </hostName>

<snip>

> > > 4. I have a template XML file and a config XML file. All the values for
> >
> > the
> >
> > > template should be available in the config file. Is that possible with
> > > libxml2?
> >
> > So you would have a template file that looks like this:
> > <hostName>
> >       <defaultValue></defaultValue>
> >       <description>
> >              <shortDescr></shortDescr>
> >              <longDescr></longDescr>
> >       </description>
> >
> >       <dataType></dataType>
> >       <operations></operations>
> >
> > </hostName>
> >
> > and, according to your example, you want to insert these values:
> > value
> > Hostname descr.
> > long descr of Hostname
> > string
> > set,delete
> >
> > into the proper places in the DOM structure.
> > Or maybe just into the proper places in a text file.
> > Or maybe as attributes on some element.
> >
> > Do I understand this correctly?
> >
> > - Rush

Here are some options:

1) If your template is really as simple as you have shown, it's probably 
easiest to just keep it as text in a std::string. Something like this:

std::string myTemplate (
"<hostName>"
"    <defaultValue>%1</defaultValue>"
"    <description>"
"        <shortDescr>%2</shortDescr>"
"        <longDescr>%3</longDescr>"
"    </description>"
"    <dataType>%4</dataType>"
"    <operations>%5</operations>"
"</hostName>");

Then, when you have assembled all of your values that you want in the 
config file, you just copy myTemplate, then use the std::string 
functions to find and replace %1 through %5. (You can wrap all this in 
an object to hide the nasty details.) Once you have the result, just 
parse it with xmlParseMemory() or xmlReadMemory() (or others?) to get 
the DOM tree. Of course, if your goal is to write the config file 
formatted as XML, then you already have the content in your string and 
you don't need libxml.

2) Parse your template file to get a xmlDocPtr for it. When you need to 
generate the config file, copy the template using xmlCopyDoc(). Now you 
have a copy of the DOM tree for the template file and you need to insert 
the element data. I can see two ways to do this:

a) Apply a XSL transformation to the tree. The downside to this is that 
you need to generate the style sheet on the fly, so that it contains the 
values that you need to insert. Again, if your template is as simple as 
your example, then the style sheet is also very simple, so is easy to 
generate. If your template file is more complicated, then the style 
sheet could get complicated as well.

b) Using one of the xmlDocDump*() functions as an example, write 
yourself a node walker and have it call back for every node. Use the 
node (element) name to decide what to do, and insert your data into the 
node structure. This only works if your template file is as simple as 
you show, and each unique element name only shows up once. If your 
template really looks like this:
<hostList>
    <hostName>
        <defaultValue></defaultValue>
        <description>
            <shortDescr></shortDescr>
            <longDescr></longDescr>
        </description>
        <dataType></dataType>
        <operations></operations>
    </hostName>
    <hostName>
        <defaultValue></defaultValue>
        <description>
            <shortDescr></shortDescr>
            <longDescr></longDescr>
        </description>
        <dataType></dataType>
        <operations></operations>
    </hostName>
</hostList>

then it gets harder, but I guess you could group your values you want to 
insert and count instances of each element, or instances of "hostName" 
to know which values you need to insert. If your template is really 
structured more like this, then this approach might be easier than using 
a XSL style sheet, because it's easier to understand how to write the 
logic in C or C++ than in XSL. It could be a maintenance nightmare too, 
if your template file format changes.

I should also mention here that I believe this is the idea behind a SAX 
interface, so you might be able to use that instead of manipulating the 
DOM. (I have never used SAX, so don't know any details. I have written a 
node walker and done in-place modifications to the DOM tree.)

3) Like 2, but don't bother to pre-parse your template because libxml 
has a really fast parser. Just parse the template file, then process the 
tree each time you need to write your config file.

There may well be other approaches. These are all I can think of right 
now. Best of luck.

Regards,
Rush