A unified theory of "no data"

["Costello, Roger L." <co...@mitre.org>]

Hello DFDL community,

I created 3 slides on ways that data formats represent "no data." See below. Your feedback on these slides would be greatly appreciated. For example, I use the term "region" ... is that the correct term?

[cid:image001.png@01D53706.FA35A9D0]


[cid:image002.png@01D53706.FA35A9D0]


[cid:image003.png@01D53706.FA35A9D0]

Re: A unified theory of "no data"

["Beckerle, Mike" <mb...@tresys.com>]

Region is good term to use. DFDL spec uses the term region in the same sense.


When you say the "region has a nil value", you are using "has" to mean "represents" which is fine most of the time.


When you say the "region is given an empty value" that is a lot more subtle, and not so clear. What is being given what? "given" here means that the element which will contain the logical value corresponding to this data region is given an empty value. This is only sensible for some types - where empty is a concept. In DFDL this is xs:string and xs:hexBinary only.


Any type, even complex types, can be nilled.

Only xs:string and xs:hexBinary can have empty values.

Any type can have an empty representation in the corresponding data region - but only if we provide a way for encountering that representation to trigger creation of a default value.

________________________________
From: Costello, Roger L. <co...@mitre.org>
Sent: Wednesday, July 10, 2019 10:05:20 AM
To: users@daffodil.apache.org
Subject: A unified theory of "no data"


Hello DFDL community,



I created 3 slides on ways that data formats represent “no data.” See below. Your feedback on these slides would be greatly appreciated. For example, I use the term “region” … is that the correct term?



[cid:image001.png@01D53706.FA35A9D0]





[cid:image002.png@01D53706.FA35A9D0]





[cid:image003.png@01D53706.FA35A9D0]