Stack Exchange hath no fury like a manager scorned

i was just thinking that someone had a poor valentine's day

@ypercubeᵀᴹ any ideas on this dba.stackexchange.com/a/164606/2639 where the LATERAL is so much slower than the case-statement method

@AaronBertrand He's got increased attention to his posts. Perhaps it's his way to "retaliate in kind". You seem like an obvious choice of the primary adversary :)

So more "I have no clue how to interact with anyone." Great.

whelp he got his first suspension.

The down-voting spree has started, too. I guess he didn't read those links I suggested after all.

Is that you cleaning up comments? Just curious - I don't know if I'm losing my mind.

Nope

On another note I wonder if Spolsky and friends are going to do another podcast soon. I want to hear his rant.

Someone else must be taking control of the situation.

Hijacking comments, you mean? I wonder who that might be.

I figure we all deserve a break

@AndriyM That is correct, that's what I had in mind (SELECT col1, x FROM foo CROSS JOIN LATERAL (VALUES (col1::text),(col2::text)) AS t(x);).

@EvanCarroll So weird that it requires ::text after a column name.

Why? SQL is strongly typed, and they're typed differently? You have to cast().

@Forrest I don't think I'm gonna make it out tonight. But ping me before the next one?

@EvanCarroll Not in the OP's case, perhaps. In their example, columns col2 and col3 that need to be unpivoted are some string type already.

Yes, the problem there is that if you're using the ops example you want to CROSS JOIN LATERAL on col2, and col3

then no casting needed.

Anyway, fuuuu that question guys

ie, (col1) in his example is the id an int.

@MaxVernon To me, the point of determining whether something is anecdotal is not that it isn't evidence, but that it is merely a single data point from which no conclusion can be drawn. Anecdotal evidence is actually evidence. But you can't draw a line from a single point. The definition you linked says systematic, which to me is the key word. Evidence has to consist of a large enough set to be statistically valid.

Well said ^.

Dawkins once had to labor this point in a debate. It was kind of amusing. I have evidence my wife loves me, you just reject it because it's anecdotal. Forever, that's how I think about anecdotal evidence. (Don't even like Dawkins)

An anecdote is simply something relayed as if it were truth, when it may or may not be. A single point of evidence can be demonstrably true, and therefore not anecdotal.

@MaxVernon I have to disagree. Something can be anecdotal and true. I don't see anecdotal as in any way implying false.

Reread what I said. I said an anecdote may or may not be true.

@MaxVernon I'm rereading, and I see "A single point of evidence can be demonstrably true, and therefore not anecdotal." You're making true and anecdotal mutually exclusive.

I think perhaps we have to distinguish between evidence and inference/conclusion/interpretation.

Sure

Evidence is simply true or false, anecdotal or no. The problem is, our inferences from anecdotal evidence are very likely to be flawed because of the fact about a single point not making a line (which is technically made up of an infinite number of points).

@MaxVernon Let's say that one time you think of John Lennon and immediately afterward, your hiccups stop. This is evidence, but is merely anecdotal, not systematic.

The point is once something is proven true, it is no longer anecdotal by definition. To be proven true, there necessarily must be more than a single data point. A conjecture can start out as an anecdote, be true, then be proven be to true by gathering none anecdotal evidence, at which point it is no longer anecdotal.

The inference that thinking of John Lennon stops hiccups is not warranted because it is based on a single data point. The evidence is true, no falseness there--it's true you did X and immediately afterward, Y happened. But you can't assign causation.

@MaxVernon I understand what you're driving at, I just object to your wording (I guess). The "something" that is being true is a different thing than the evidence. I would correct your last statement this way:

@MaxVernon Wait, I can't. The something that's proven isn't the evidence, it's the conjecture.

@MaxVernon "The point is once a conjecture is proven true, the conjecture is no longer anecdotal by definition" doesn't make sense. And "the point is once a conjecture is proven true, the evidence it's based on is no longer anecdotal by definition" also doesn't make sense to me.

why aren't you both upvoting all of my questions and answers?!

@MaxVernon One piece of anecdotal evidence isn't enough. 20 may not be enough. But 100 or 1000 may be enough. At that point, the body of evidence is called systematic and we can start to draw causational or at least correlative conclusions. It's like the heap conundrum (the open/closed concept). A single grain of sand isn't a heap, but 10,000 of them is. Where is the cutoff where n grains of sand is not a heap but n+1 is?

@MaxVernon The amount of evidence required before it can be considered systematic evidence instead of anecdotal isn't fixed or always easily determined. In some cases it might only take 5 points, in others it might take thousands or millions, depending on how subtle the difference is that we're trying to analyze and our understanding of the context of the problem already.

@MaxVernon At the risk of alienating you due to too many words (sorry)... if I prove beyond a shadow of a doubt that first I thought of John Lennon and immediately after my hiccups stopped, this proving that this evidence is true (it actually occurred) doesn't make it non-anecdotal. That is a separate thing from the assignment of causation from the first to the second, our conjecture.

@AaronBertrand Can you, Aaron (or someone), point me to a link describing your position on avoiding MERGE? I have been using it but only ON 1 = 0 and ON UNMATCHED INSERT so I can OUTPUT one of the source table's columns along with some INSERTED.Column values.