OASIS Mailing List ArchivesView the OASIS mailing list archive below
or browse/search using MarkMail.

 


Help: OASIS Mailing Lists Help | MarkMail Help
[Date Prev][Date Next][Thread Prev][Thread Next][Date Index][Thread Index]

Re: [xml-dev] storing XML files

Albena Georgieva wrote:
> Ronald: What do you mean by "semi-structured"_ness of the data?

Structured data is very rigid. That is, all records have the same
fields. An example of structured data is a telephone book -- every entry
has a name, an address, and a telephone number.

Semi-structured data is data that has some structure, but is not rigidly
structured. An example of semi-structured data is a health record. For
example, one patient might have a list of vaccinations, another might
have height and weight, another might have a list of operations they
have had. Other examples of semi-structured data are legal documents and
geneological records.

A common example in industry is data that has been retrieved from many
different sources. For example, if you ask for all the data about
customer X, you might get a sales history, emails, stock profile, and so
on. However, the set of data will differ greatly from customer to
customer.

Semi-structured data is difficult to store in a relational database
because it means you either have many different tables (which means many
joins and slow retrieval time) or a single table with many null columns.
Semi-structured data is very easy to store as XML and is a good fit for
a native XML database.

> <!--     EXAMPLE -->
> 
> <?xml version="1.0" encoding="ISO-8859-1" ?>
> <!DOCTYPE nitf SYSTEM "nitf-adjusted-13c.dtd">
> <nitf>
> <head>
> <meta name="onlinefolder" content="Entertainment"></meta>
> <docdata>
> <date.issue norm="20010302 150629+0100"></date.issue>
> <key-list>
>                 <keyword key="Maxima"></keyword>
>                 <keyword key="Princess"></keyword>
>         </key-list>
> </docdata>
> </head>
> <body>
> <body.head>
> <hedline><hl1>Maxima speaks perfect dutch</hl1></hedline>
> </body.head>
> <body.content>
> <p>Princess Maxima presented the Princess Collection 2000 in an estate with
> a very royal ambiance. The dazzling collection is also presented in the new
> catalogue full of show, glitter and glamour.</p>
> </body.content>
> </body>
> </nitf>
> 
> <!--     /EXAMPLE -->
> 
> All I want to do with them for now, is to save them in a RDBMS. The rest of
> the applications (ASPs or servlets) will access that data through ODBC or
> JDBC.
...
> If the applications access the datafeeds through ODBC or JDBC and they don't
> ask for XML format, ( no need for XML retrieval ) I see no reason for
> introducing a XML native database at this point, but please correct me I am
> wrong ...

These two requirements -- storing NITF in a relational database and not
returning any XML -- conflict with each other. If you don't want to
return XML, that means no XML can be stored in the database. But if you
map NITF to a relational database using the most widely accepted mapping
that doesn't store XML in the database (an object relational mapping),
you will end up with an almost useless set of tables.

The problem is that NITF uses mixed content and mixed content doesn't
map well with an object-relational mapping. (I won't go into the details
here. If you want to read more about this, see sections 3.3 and 3.4 of
http://www.xml.com/pub/a/2001/05/09/dtdtodbs.html.)

Furthermore, as far as I know, the only product that supports mixed
content in an object-relational mapping is mine (XML-DBMS) and I
recommend that people don't use XML-DBMS with mixed content because it
is so inefficient. (XML-DBMS started out as a research project and I
supported mixed content for completeness. Were I to do it over again, I
probably wouldn't support it.)

Your choices therefore are:

1) Store your documents in a native XML database.

2) Store your documents as BLOBs in a relational database and index them
as Soumitra suggests. (For an example of this technique, see the
discussion of side tables in section 6.2 of
http://www.rpbourret.com/xml/XMLAndDatabases.htm.)

3) Store the documents in a relational database with object-relational
middleware. However, instead of storing mixed content in multiple
tables, store it as XML in a single column. For example, you would store
the content of the <body.content> element in a single column as:

   <p>Princess Maxima presented the Princess Collection 2000
   in an estate with a very royal ambiance. The dazzling
   collection is also presented in the new catalogue full
   of show, glitter and glamour.</p>

Notice the <p> elements. Not all middleware products support this
technique, but some do. Another problem with this choice is that it is
not clear if non-mixed-content parts of the NITF DTD can even be mapped
to a useful set of tables. With some transformations, it might be
possible, but I'm not sure.

In any case, your applications are almost certainly going to have to
deal with some XML. Otherwise, there doesn't seem to be any way to deal
with the mixed content.

> I just need a good way to transform a XML data feeds into relational
> database model. That is why, I asked for something like DTD2SQL or XML2SQL
> tool. So, I imagined applying XSLT transformations to every different
> datafeed to transform it into some DTD for SQL and then just run that SQL.

This is possible. That is, you could write an XSLT transformation that
transformed your documents into a series of tables. For example,
transform a document with the form:

   <table1.row>
      <table1.column1>...</table1.column1>
      <table1.column2>...</table1.column2>
      <table2.row>
         <table2.column1>...</table2.column1>
         <table2.column2>...</table2.column2>
         <table2.column3>...</table2.column3>
      </table2.row>
   </table1.row>

to the form:
     
   <table1.row>
      <table1.column1>...</table1.column1>
      <table1.column2>...</table1.column2>
   </table1.row>
   <table2.row table1.key="...">
      <table2.column1>...</table2.column1>
      <table2.column2>...</table2.column2>
      <table2.column3>...</table2.column3>
   </table2.row>

It would then be easy to insert these rows into tables. (Note that there
is no need for a standard DTD here. The correspondence of elements to
tables and columns is inherent in the structure. To be truly generic,
all that would be needed is a simple mapping file that mapped element
and attribute names to table and column names.)

The data transfer code is easy to write, although the stylesheets
probably aren't. (I'm actually surprised that nobody has written a
product that takes an XML-to-DBMS map and generates the above
stylesheets. It wouldn't be the most efficient way to transfer data due
to the XSLT transform, but it has the advantage of simplicity.)

In any case, the code and the stylesheets aren't the problem. The
problem is the one mentioned above -- that the NITF DTD simply doesn't
map well to relational schema due to its mixed content.

-- Ron