SQL:2023 says SQL is “pseudo-relational” instead of truly relational.
The (free) Part 1 of the ISO SQL:2023 standard, under the heading "What is SQL?" on page 6, says (my emphasis):
SQL ... is a database language ... used for access to pseudo-relational databases that are managed by pseudo-relational database management systems (RDBMS).
and it goes on to explain that
SQL is based on, but is not a strict implementation of, the relational model of data, making SQL “pseudo-relational” instead of truly relational.
and it elaborates
- The relational model requires that every relation have no duplicate rows. SQL does not enforce this requirement,
- The relational model does not specify or recognize any sort of flag or other marker that represents unspecified, unknown, or otherwise missing data values. Consequently, the relational model depends only on two-valued (true/false) logic. SQL provides a “null value” that serves this purpose. In support of null values, SQL also depends on three-valued (true/false/unknown) logic.
Are those words new in the 2023 version, or were they there before?
Also do we know who proposed those words to be added?
Words. Last I looked SQL does fully support the RM. It also supports an Extended RA, although it's hard to find agreement on what that means. IOW there is no feature of an RM or ERA that is beyond what SQL can do.
But it messes things up by including a lot of other crap as well. And the SQL type system is a load of hokum.
BTW there is also this footnote on the same page
1 Terms defined by the relational model are used correctly in this Subclause, but are not defined or explained in this document; readers are referred to the formal specification of the relational model
I.e. the SQL Standard is not brave enough to define or formally specify the relational model itself
The page simply has Relational Model as a hyper link that refers to the first of the 2 items in it's Bibliography:
[1] Codd, E. F. A Relational Model of Data for Large Shared Data Banks [paper]. New York: Association of Computing Machinery, June 1970.
I have a strong distrust of reliance on ancient papers. Codd's RM is a good start, but the RA falls well short of what's needed in a modern database. It has no type system, negation or recursion quite apart from the other things mentioned above.
[If anyone knows a good paper on extended RA I'd be glad to hear of it.]
Quote from dandl on March 26, 2025, 10:56 pm... a good paper on extended RA ...
emm D&D's writings? Or perhaps I should ask: just what do you mean by "extended RA", seeing as how there are about a gazillion ways people claim to have extended RA-alikes. So I think capitalising 'Extended' per your first post is not merited.
Quote from Paul Vernon on March 26, 2025, 11:57 am
Are those words new in the 2023 version, or were they there before?
I'd say describing SQL as 'pseudo-relational' would date back to soon after SQL was released/people realised what a hash IBM had made of the idea. And note IBM and others rapidly produced several implementations that were closer.
For example in a ycombinator discussion 2010 arising from a javaworld article about ORM. Or on the C2 wiki, coincidently the same year, in which Dave participated.
As to author, it sounds like the style of Andrew Warden. Hugh was on the SQL Committee for many years. Wikip alleges (follow the links from there) older versions of the standard are downloadable from that same ISO site you link to. Their search tool is what I'd call user-hostile: I gave up trying.
BTW, the other day I fell across Javascript's 'Falsy and Truthy'. WTF? I didn't think it was possible to invent something worse than SQL's NULL. Fans of obfuscated code will be delighted to hear NaN is falsy; they'll be rubbing their hands in glee at the prospect of exploiting the semantics of the Booly operators && and ||.
Quote from AntC on March 27, 2025, 1:27 amQuote from dandl on March 26, 2025, 10:56 pm... a good paper on extended RA ...
emm D&D's writings? Or perhaps I should ask: just what do you mean by "extended RA", seeing as how there are about a gazillion ways people claim to have extended RA-alikes. So I think capitalising 'Extended' per your first post is not merited.
I mean one that adds significant capability beyond that in Codd, and ultimately one that can express every query possible in SQL. If you recall Alice, I mean at least SPJRUN, but ultimately including recursion and and ordered queries. I'm not aware of a single example of this being done as a formal definition or a programming language, but Andl gets very close.
BTW, the other day I fell across Javascript's 'Falsy and Truthy'. WTF? I didn't think it was possible to invent something worse than SQL's
NULL. Fans of obfuscated code will be delighted to hearNaNis falsy; they'll be rubbing their hands in glee at the prospect of exploiting the semantics of the Booly operators&&and||.
You do lead a sheltered life. Yes, JS is worse than SQL (although it's a close race), but equally pervasive and useful in practice. Most people who care move on to Typescript or other JS variants or cross-compilers. I'm currently maintaining a 43KLOC project in JS, and I'm telling you, it bites!
And I should have mentioned, outside of the ERA itself you need
- database variables and assignment (database variables may be local or server, but the code is the same).
- some form of expression evaluation in the attribute type system, possibly based on Algebra A relational constants, which leads to another discussion entirely.
Quote from dandl on March 27, 2025, 4:32 amI mean ... every query possible in SQL.
Oh dear. I fear this'll be a portal into another of those circular terminological debates. Look at the very first post in this thread. By SQL (2023[**])'s own admission
> ... making SQL “pseudo-relational” instead of truly relational.
You can have Relational or you can have SQL. Pick one. I'd add: you can have an Algebra or you can have SQL SELECT. Pick one. Also running "formal definition" and "programming language" in the same sentence seems like a category mistake.
[**] ISO is extraordinarily closely-guarded. I've been down a lot of rabbit-holes claiming to show me older copies of ISO-9075 Part 1, but they keep leading to the 2023 wording, or wanting to charge big bucks. This appears to be a draft 1992 standard, but beware the caveats at the top of the doco. "Pseudo-" anything doesn't appear. Indeed the doco neither asks nor answers "What is SQL?" Neither "relational" nor "relation" appears. "Model" appears only as in "data model" of some example schemas, not 'XXX model of data'. "Codd" doesn't appear.
Quote from AntC on March 27, 2025, 5:37 amQuote from dandl on March 27, 2025, 4:32 amI mean ... every query possible in SQL.
You can have Relational or you can have SQL. Pick one. I'd add: you can have an Algebra or you can have SQL
SELECT. Pick one. Also running "formal definition" and "programming language" in the same sentence seems like a category mistake.
I don't think you're paying attention. The choice is relational, with the capabilities to match and superset SQL queries. Algebra, not SELECT. Doesn't seem complicated to me, and Andl already does most of it.
And 'formal definition' in the same sense as IEEE754 or Unicode, which a programming language may implement and use.
And a separate type system, somewhat like JSON. Values, not OO, so most widely used languages are out.
Regarding falsiness and truthiness, it's proponents would claim it is obvious, but none of the languages using the concept agree