Question

I have some unstructured data that describe setting for each devices. For e.g. for Device 1:

<ChartSetting xmlns="ChartSetting1">
  <ChartSizeX>132.6</ChartSizeX>
  <ChartSizeY>132.6</ChartSizeY>
  <ChartType>Native</ChartType>
  <BarSizeX>90</BarSizeX>
  <BarSizeY>6</BarSizeY>
  <DistToBar>34.8</DistToBar>
  <DistToFirstLineY>17.5</DistToFirstLineY>
  <MarkerDistance>120</MarkerDistance>
  <DistToFirstField>18.5</DistToFirstField>
  <PatchSizeX>7.5</PatchSizeX>
  <PatchSizeY>9</PatchSizeY>
</ChartSetting>

However for Device 2 setting is different

<ChartSetting xmlns="ChartSetting2">
  <PatchGap>1</PatchGap>
  <PatchSize>5</PatchSize> 
</ChartSetting>

xml data is not used for query purpose, it will be passed to the .net application that eventually send it to the devicedriver (through C++ code)
We only have four types device. So possibility of different settings is limited.

1. Can this kind of data be stored as typed XML that adhere to some schema? Storing as varchar will be nightmare if there are invalid settings stored.

   I know xml schema collection can have multiple schema but can it confirms to only one schema.

2. Is there an easy way to create a schema?

3. Should I be using untyped XML instead?

Answer 1

1.I think it’s possible, if your schema collection contains 4 different schemas, the related column can contain xml that satisfies one of the schemas in collection.
2.I believe the easiest way is to let sql server create a schema for you (first create tables ChartSetting1,ChartSetting2 with desired columns – it’s not necessarily, you can create SELECT that returns all columns, but using tables is easier), then

DECLARE @mySchema NVARCHAR(MAX);  
SET @mySchema = N'';  
SET @mySchema =@mySchema  +   
(
  SELECT * FROM ChartSetting1 
  FOR XML AUTO, ELEMENTS, XMLSCHEMA('ChartSetting1')
);

SET @mySchema =@mySchema  +   
(
  SELECT * FROM ChartSetting2 
  FOR XML AUTO, ELEMENTS, XMLSCHEMA('ChartSetting2')
);

CREATE XML SCHEMA COLLECTION [name] AS @mySchema;

-- we don't need these tables , drop them  
DROP TABLE ChartSetting1,ChartSetting2

3.It depends. I’d say if you need constrain the XML data beyond schema validation, then it makes sense to use untyped XML. If schema validation covers all aspects of data validation, why not use it?