Was: Inconsistent behavior of `sum` and `mean`


This seems overly harsh to me. Scott clearly presented an inconsistent behaviour: sum([foo,bar]) vs sum((foo,bar)) – that is unarguably undesirable. This may not match the dictionary definition of “bug” or “broken”, but the basic sentiment is surely correct… If anyone is being offended by these words then meh, is this not being a little hypersensitive / combative even, and taking offence where none has been given (let alone intended)?

Also the original title has been changed for something that fails to convey the original meaning. At least “Inconsistent overflow behaviour…” surely?? The central point of his post was to draw attention to the inconsistency, and changing the title thus does no more than dull the effect. I would be rather miffed if someone reworded a post of mine. My words on my own!

I don’t see anything worthy of rebuke in what he has posted – it appears to be a useful community contribution. Hence I can only surmise from the wording of “A tip that you’ve heard before” that there is some kind of grudge from a previous time.

I feel it is important to give someone the benefit of the doubt; today is a fresh day, let the sounds of yesterday fade away – the nectar is in the moment!

As for “raising multiple independent issues”, I’ve just been told off for that too. My bad I’m sure (I should remain quiet on technical issues, I have little to contribute). But the dictionary definition of “discourse” is “open discussion”, and is that not exactly what we want here as a community? It sounds as if the same stringency of protocol that applies on the Github Issue Tracker is being applied here. I thought the whole point of Discourse was a more relaxed/open-ended forum for discussion. Otherwise what is the point in its existence? If this is just going to operate in the same manner as the Issue Tracker… Isn’t that just creating a headache by splitting issues into 2 places?

Doesn’t it make more sense that this place is used for (more) general pow-wow and the fruits of the discussion make it to the Github issue tracker, which can then be kept nice and clean?

It seems a sensible and natural ordering: Gitter (for the most general and most transient conversation), Issue Tracker (for focused permanent conversation) and Discourse in the middle.

I feel I’m walking a fine line in speaking my mind. I sure wish someone else had spoken up instead. I need the Julia community much more than it needs me, and writing this certainly doesn’t serve my own interests a jot.

Inconsistent behavior of `sum`,`mean` (and probably others) on different collection types

I’d ask you to consider the possibility that you’re also not serving the interests of Julia as an open source project. I personally believe that every single committer to Julia wants to provide an inclusive and welcoming environment. But “open discussion” doesn’t put in N hours of work per week to develop and maintain Julia: specific people do. If the rules of open discussion create an environment that those specific people find to be unpleasant, you’ll achieve open discussion at the cost of the personal well-being of the people who keep Julia growing. I don’t think that’s a good trade-off to make.


I have moved these posts to a new thread under #community since they are no longer about the original technical topic. Discussion can continue here or not.

This seems overly harsh to me.

@pitao: without more context, I would agree with you, but there’s a lot of history here. Specifically there is a history of @ScottPJones raising and pursuing many vaguely connected issues all at once, and implying or outright asserting that project maintainers have implemented things incorrectly in cases where there is simply a difference of opinion over how things should work. Frankly, the community has still not entirely recovered from the unpleasant environment that this chronic behavior caused – in particular our willingness to discuss all matters in an open forum has never bounced back to its former level. Most core contributors are still more than just a little touchy when presented with even a small amount of that familiar problematic behavior.


@StefanKarpinski I can see there is no easy answer… these vexing situations are a challenge to conscious action. I was concerned that maybe StevenGJ was reacting to something in the past that may not be here in the moment, and that the reaction may provoke a counterreaction and summon old grievances from their graves.

@johnmyleswhite While I can see the truth in everything you write, I can’t quite see how it pertains to my post. I am continually considering the consequences of my actions and on account of this write with some trepidation. I don’t want to be chewing up a single moment of developer time. You talk about an “unpleasant environment”, but on this occasion has this not been created today by one developer taking a bit of a snipe at another? I’m sure it can be justified, but isn’t justification the sustenance of strife?

It looks as though Scott must have really whacked the hive in the past, nevertheless I must have some faith in a man’s ability to change up (given sufficiently many bee-stings).


@pitao: Please not that Julia has developed Community Standards: http://julialang.org/community/standards/
and Steven is one of the stewards. There is a certain level of trust that has to be given to them. But as we have seen in the past it is really important to have such an institution.


To my reading, that post wasn’t very clear, at least not initially. I believe the title was changed in good faith as an attempt to sharpen the discussion on a particular behavior that was being considered buggy. I don’t see a request to keep discussion concise and focused a “snipe”. And, yes, past behavior does color how things are read. If someone regularly uses the word “bug” to talk about a missing feature, then it’s hard to know how they’re using the word in the future.

But here’s the most important thing: neither part of this is contentious on a technical level. Within the first two replies it was acknowledged that, yes, overflow behaviors can be wonky, but it’s a deliberate design choice. And yes, we can make the promotion behaviors more consistent. There’s no need to belabor the point or argue about the severity thereof.