Hi everyone.
I've been asked by Alok of the AI team to consider replacing
empty-string values in XML nodes with a default.
Currently, defaults are added only when nodes are missing from under a
parent node; this infers that if a node is present, even if it has an
empty string for a value, its value is considered to be present and
valid. A parent's child node is either there or it isn't.
There are times where one might want to replace an empty-string value
with a different one. For example, the AI project may wish to specify
something like this:
<device></device>
in a manifest and have defaults processing fill in the value. This not
only documents with a placeholder that a certain element is specifiable
in the manifest, but it allows for easy automation of filling in the
value. This improves upon having a commented value or comment for a
placeholder, because it isn't as easy to uncomment something in
automated fashion as it is to just fill in a value.
There may also be times where an empty string should be flagged as an error.
When assigning defaults to blank valued nodes, what about multiple
sibling nodes with the same nodepath? For example, what to do if a
parent has two children (which match the same nodepath), and one or both
are blank? I propose that all blanks get assigned the same default
value. It's easiest to implement this way, and in cases where this
doesn't make sense, there shouldn't be multiple blank values in the
first place.
I propose that:
1) Each "default" entry in the defval manifest have a new optional tag:
empty_str. It could be set to
"valid", "replace" or "error". If set to "valid", empty strings would
be assumed OK as is, and no default assigned to replace them. If
"replace", all empty string values would be replaced by a default
value. If "error", well, you know the drill...
2) If a parent node has no child nodes matching the nodepath, one new
child node with a default value will be provided. (Same as before)
3) If a parent has one or more child nodes which match the nodepath and
have an empty string, the value of those nodes would either be skipped,
assigned the same default value, or flag an error, depending on the
"empty_str" tag assigned to their nodepath in the defval manifest. This
would happen regardless of whether the parent has any non-blank child
nodes which match the nodepath.
I'm not sure that there's time to get this in for November, but I'd like
to take your comments on this proposal. Please send your feedback by
Monday 10/6 COB.
Thanks,
Jack
