Schema for simple XML-based scripting language

[Chris Lieb]

I am new to XML Schema and am running into a bit of a snag. I have
defined an XML-based scripting language for an updater program that I
am working on. I would like to make a schema for this language since
malformed XML documents break the updater. (I did not have time to add
good error-handling code, so run-time errors can abound if the document
is malformed.)

A sample doc might look something like this:

<manifest>
<version major="1" minor="0" revision="1" base="1.0.1\files\"
doprevious="true">
<file action="copy" name="text.doc"/>
<folder action="create" name="temp"/>
<updater>
<file name="update.exe"/>
<file name="updaterlib.dll"/>
</updater>
</version>
<version major="1" minor="0" revision="0" base="1.0.0\files\"
doprevious="false">
<dll action="register" name="updaterlib.dll"/>
<cmd location="server" command="patch.bat">
<file name="patch.dat"/>
<file name="patch.bat"/>
</cmd>
</version>
</manifest>

Problem is that the second version element in the document, the updater
element, the dll element, and the cmd element all get flagged as
invalid. I guess that this stems from using a sequence in the complex
type. What should I use to allow all children elements of version
(file, folder, dll, msi, cmd, updater) to appear in any order and in
whatever quantity (except updater, which can only appear once)?

'all' looks like it comes close since it allows you to use elements in
any order, except my reading tells me that it only allows each element
to be used once and forces you to use all elements. 'choice' also
appears to come close since it allows you to choose which element you
want, but it only lets you choose one element from the set.

Also, one last thing: How do you set the max occurrence of an element
to be infinite?

Thanks in advance,
Chris

PS
I'm not posting the XSD right now. If you want it, I would be more
than happy to provide it. I am a little embarassed that I am designing
the XSD in Visual Studio 2005 instead of hand-coding it. I have no
idea if the XSD generated by VS2005 is considered to be of good
quality.

[Chris Lieb]

J Hendrich wrote:

"Chris Lieb" wrote:

I am new to XML Schema and am running into a bit of a snag. I have
defined an XML-based scripting language for an updater program that I
am working on. I would like to make a schema for this language since
malformed XML documents break the updater. (I did not have time to add
good error-handling code, so run-time errors can abound if the document
is malformed.)

A sample doc might look something like this:

<manifest>
<version major="1" minor="0" revision="1" base="1.0.1\files\"
doprevious="true">
<file action="copy" name="text.doc"/>
<folder action="create" name="temp"/>
<updater>
<file name="update.exe"/>
<file name="updaterlib.dll"/>
</updater>
</version>
<version major="1" minor="0" revision="0" base="1.0.0\files\"
doprevious="false">
<dll action="register" name="updaterlib.dll"/>
<cmd location="server" command="patch.bat">
<file name="patch.dat"/>
<file name="patch.bat"/>
</cmd>
</version>
</manifest>

Problem is that the second version element in the document, the updater
element, the dll element, and the cmd element all get flagged as
invalid. I guess that this stems from using a sequence in the complex
type. What should I use to allow all children elements of version
(file, folder, dll, msi, cmd, updater) to appear in any order and in
whatever quantity (except updater, which can only appear once)?

'all' looks like it comes close since it allows you to use elements in
any order, except my reading tells me that it only allows each element
to be used once and forces you to use all elements. 'choice' also
appears to come close since it allows you to choose which element you
want, but it only lets you choose one element from the set.

Are you certain you need to allow for all possible combinations of all
elements in any order...if you can whittle down the possible combinations to
only those combinations that should be allowed to occur you can use
xs:sequence embedded inside of xs:choice, like:

<xs:choice>
<xs:sequence>
<xs:element ref="dll"/>
<xs:element ref="cmd"/>
</xs:sequence>
<xs:sequence>
<xs:element ref="file"/>
<xs:element ref="folder"/>
<xs:element ref="updater"/>
</xs:sequence>
</xs:choice>

If indeed all possible combinations are possible, then the above is not a
very attractive solution.

Also, one last thing: How do you set the max occurrence of an element
to be infinite?

maxOccurs="unbounded"

PS
I'm not posting the XSD right now. If you want it, I would be more
than happy to provide it. I am a little embarassed that I am designing
the XSD in Visual Studio 2005 instead of hand-coding it. I have no
idea if the XSD generated by VS2005 is considered to be of good
quality.

I've used a few different xsd generators...I'd suggest getting familiar
enough with XML Schema to know whether or not you like what it is generating.
Each generator makes its own assumptions about certain things, and you won't
always like what it assumes. I have only tried the XSD in VS 2003, and I
don't know if it applies to the one in 2005, but the 2003 version cannot
generate schemas for a broad class of XML files (whereas other generators
don't have a problem). I found that to be an annoying limitation and have
stayed away since.

I would like to allow items in any order in any quantity. I feel that
it might restrict thing in the future if I force you to always define
folders before files and so forth. The only restriction that I want to
place on the order is to force the <updater> node (max of 1, min of 0)
to always be the last item for a version.

I guess that I might not actually need a schema for the manifest files
since I am making a GUI to allow others to easily edit the manifest
file without needing to know all of the rules that I have or even XML.
Even if they did know XML, I am not currently validating it in the
updater using XML Schema, but rather by manually checking values.

I could make a change to the updater to force validation against a
schema, but I don't know if MSXML4 supports this, or how to access it
through VB6. Any ideas?

Do you think that a schema would actually be useful in this case?

Chris

[Chris Lieb]

Chris Lieb wrote:

J Hendrich wrote:

"Chris Lieb" wrote:

I am new to XML Schema and am running into a bit of a snag. I have
defined an XML-based scripting language for an updater program that I
am working on. I would like to make a schema for this language since
malformed XML documents break the updater. (I did not have time to add
good error-handling code, so run-time errors can abound if the document
is malformed.)

A sample doc might look something like this:

<manifest>
<version major="1" minor="0" revision="1" base="1.0.1\files\"
doprevious="true">
<file action="copy" name="text.doc"/>
<folder action="create" name="temp"/>
<updater>
<file name="update.exe"/>
<file name="updaterlib.dll"/>
</updater>
</version>
<version major="1" minor="0" revision="0" base="1.0.0\files\"
doprevious="false">
<dll action="register" name="updaterlib.dll"/>
<cmd location="server" command="patch.bat">
<file name="patch.dat"/>
<file name="patch.bat"/>
</cmd>
</version>
</manifest>

Problem is that the second version element in the document, the updater
element, the dll element, and the cmd element all get flagged as
invalid. I guess that this stems from using a sequence in the complex
type. What should I use to allow all children elements of version
(file, folder, dll, msi, cmd, updater) to appear in any order and in
whatever quantity (except updater, which can only appear once)?

'all' looks like it comes close since it allows you to use elements in
any order, except my reading tells me that it only allows each element
to be used once and forces you to use all elements. 'choice' also
appears to come close since it allows you to choose which element you
want, but it only lets you choose one element from the set.

Are you certain you need to allow for all possible combinations of all
elements in any order...if you can whittle down the possible combinations to
only those combinations that should be allowed to occur you can use
xs:sequence embedded inside of xs:choice, like:

<xs:choice>
<xs:sequence>
<xs:element ref="dll"/>
<xs:element ref="cmd"/>
</xs:sequence>
<xs:sequence>
<xs:element ref="file"/>
<xs:element ref="folder"/>
<xs:element ref="updater"/>
</xs:sequence>
</xs:choice>

If indeed all possible combinations are possible, then the above is not a
very attractive solution.

Also, one last thing: How do you set the max occurrence of an element
to be infinite?

maxOccurs="unbounded"

PS
I'm not posting the XSD right now. If you want it, I would be more
than happy to provide it. I am a little embarassed that I am designing
the XSD in Visual Studio 2005 instead of hand-coding it. I have no
idea if the XSD generated by VS2005 is considered to be of good
quality.

I've used a few different xsd generators...I'd suggest getting familiar
enough with XML Schema to know whether or not you like what it is generating.
Each generator makes its own assumptions about certain things, and you won't
always like what it assumes. I have only tried the XSD in VS 2003, and I
don't know if it applies to the one in 2005, but the 2003 version cannot
generate schemas for a broad class of XML files (whereas other generators
don't have a problem). I found that to be an annoying limitation and have
stayed away since.

I would like to allow items in any order in any quantity. I feel that
it might restrict thing in the future if I force you to always define
folders before files and so forth. The only restriction that I want to
place on the order is to force the <updater> node (max of 1, min of 0)
to always be the last item for a version.

I guess that I might not actually need a schema for the manifest files
since I am making a GUI to allow others to easily edit the manifest
file without needing to know all of the rules that I have or even XML.
Even if they did know XML, I am not currently validating it in the
updater using XML Schema, but rather by manually checking values.

I could make a change to the updater to force validation against a
schema, but I don't know if MSXML4 supports this, or how to access it
through VB6. Any ideas?

Do you think that a schema would actually be useful in this case?

Chris

Ok, I got it mostly working. The elements (except <updater>) can
appear in any order in any quantity and <updater> is forced to be last
if it exists.

One last problem. The <cmd> element has a location property that can
be either 'server' or 'client'. If it is client, the contents of the
command attribute are run directly on the command line. If it is
server, I process the <file> and <folder> elements that are children of
the <cmd> element, execute the contents of the command attribute from
the application's temp directory, and then delete all of the files
created by the <file> and <folder> elements that are children of the
<cmd> node.

Is it possible to define an element that must have (greater than 1)
children of certain types if it has an attribute with one value, but
have zero children if it has an attribute with another value? If so,
how?

Chris

[Chris Lieb]

J Hendrich wrote:

I guess that I might not actually need a schema for the manifest files
since I am making a GUI to allow others to easily edit the manifest
file without needing to know all of the rules that I have or even XML.
Even if they did know XML, I am not currently validating it in the
updater using XML Schema, but rather by manually checking values.

I could make a change to the updater to force validation against a
schema, but I don't know if MSXML4 supports this, or how to access it
through VB6. Any ideas?

Do you think that a schema would actually be useful in this case?

Chris

If you're only using the schema to validate that your code produces a valid
manifest file (based on the user's input to your gui) that's still useful
(IMHO). You don't have to put it in your application's logic to validate
every file since the creation of the manifest file is under your
control...but if you do have it there, you might look at it like having
test/debug code to catch errors early on.

Part of the reason that I want to build validation logic into the
updater client is that I am making a quick and dirty GUI for editing
the manifest since I have four projects going and have to have them
finished by the end of June and this one is of lower priority. (That's
the problem with managers that don't understand that it is a waste of
the their and the devs' time to have to manually install every update
every time that a breeakage is encountered, which is often in this
beta-testing phase.) If I knew that I would have a decent amount of
time to work on the editor, I might forgo the schema.

As for splitting the node type, I considered it, but I don't want to
make the dialect more complicated than it needs to be :) (oops!). I
guess that splitting the node type might simplify the dialect, but I
don't have enough experience to really know.

Chris