Not pass by (no functions in this article) but it's basically the same principle for types and assigning objects to new variables- still just a reference
Well yea, you are kind of comparing apples with pears.
The same thing would happen without the function. For the reference to not break, you cannot reassign p - only it's properties.
leta={v:1};constfoo=(p)=>{p.v=2};foo(a);console.log(a);// prints { v: 2 }
.. this would still be a reference and mutate a.
Same gist, but without the function
letc={someText:"Hello"}letd=c;d={v:2}console.log(c);// prints { someText: "Hello" }
In this code it acts like you passed the memory location foo(#234)
when you change it with p = { v: 2 }; you're creating entire new object { v: 2 } this objects has a new memory location like #427, you're changing value #234 with #427 so a is also changing
Edit: value of a is not changing
leta={v:1}functionchangeValue(x){x={v:2}// you're not really changing value of a you're changing value of x which is reference to value of a }changeValue(a)a// { v: 1 }
a changes when you change property of a in function
leta={v:1}functionchangeProp(x){x.v=2// a is also changing because by changing property you're not changing reference to `{ v: 1 }`, you're changing value inside that reference}changeProp(a)a// { v: 2 }
Again, you reassign the variable. That is indeed pass-by-value but only because you reassign the variable, which in any case you shouldn't do because it's semantically confusing.
Have a look at the style guide by AirBnB for example on this topic (github.com/airbnb/javascript#es6-d...)
In other words (not mine):
if your value is an object, then the new scope gets a copy of the reference. If you modify the object by dereferencing (that is, a.foo = "bar";), your modifications to the object persist outside of your scope.
Your change does not persist outside of the function-scope because you did not mutate the original reference to a. You created a "temporary" version of it inside the function, which doesn't exist anymore because of the function scope. If you want the change to persist (pass-by-ref), you have to dereference it (look at my earlier example).
Coding since 11yo, that makes it over 30 years now ~~~
Have a PhD in Comp Sci ~~~
Love to go on bike tours ~~~
I try to stay as generalist as I can in this crazy wide place coding is at now.
consta=somethingfunctionfoo(value){codez}foo(a)// <= this could never change the type of a, since foo never sees a itself, just its valuefunctionbar(pram){pram.baz=1// <= this isn't mutating pram, but the details of the object pram refers to.pram=1// <= this mutates pram itself. Were it pass-by-reference then the argument to the function in the caller scope would now read as 1, because it's the thing we just set to 1}
But, the "value" of a variable for a JS object is a reference to the underlying object.
This is the same with strings, but they're immutable so their references cause less confusion
consta="hi"constb="hi"// a and b are two values that refer to the one common string in memorya="bye"// reassigning a doesn't change b
So, functions pass by value, but in many cases that value is just a reference to some common structure.
Hey man, just some feedback on your conversational style:
I find it's quite confrontational and irritating to read.
To me it feels like your trying to score internet points for being right and aren't actually interested in explaining things to people.
It reads more like a Twitter or Reddit thread and I was always very happy to see a calmer more constructive and friendly style on here.
I really don't mean this as a personal attack, I just feel the conversation could have a lot more merit to others (especially newcomers) if it were more constructive.
Coding since 11yo, that makes it over 30 years now ~~~
Have a PhD in Comp Sci ~~~
Love to go on bike tours ~~~
I try to stay as generalist as I can in this crazy wide place coding is at now.
I'm glad we can all agree that javascript doesn't support pass by reference.
Now, could you show me where in the ecmascript standard it talks about an object value being a reference?
vvvvvv
(First para deleted)
What do you mean that object values are references? I've been fishing through the spec (ecma-international.org/ecma-262/11...) and can't find anything saying so.
Same info and questions, but the second one is showing you're willing to put in some work to find your answer, that we probably fundamentally agree but are stuck on terminology, and will ultimately avoid nag replies like this.
If the language doesn't come naturally to you, try your best to fake it til you make it.
I had a squiz at the spec, couldn't find a specific mention, and ran out of motivation to dig further. The proof though is
a={}b=aa.foo=1console.log(b.foo)// <- prints "1"
For a time, the values of a and brefer to the same object. They hold two pointer-like structures that reference the same shared underlying object. In programming this is often called a reference.
Coding since 11yo, that makes it over 30 years now ~~~
Have a PhD in Comp Sci ~~~
Love to go on bike tours ~~~
I try to stay as generalist as I can in this crazy wide place coding is at now.
That's fine, but for future reference, let's note that a.foo doesn't mean that a is a reference.
It does mean that a and b have the same value, and it does mean that that value is used to find a particular property named 'foo' and modify that.
Which makes me think that all of this 'reference' stuff is something someone made up to try to explain javascript in terms of a language with references, like C++, but which doesn't quite fit.
Wow, for some reason, reading on mobile, I totally missed the article, mixing it up with @alexanderjanke
's comment! (so, my comment was actually discussing the GIF, which depicts passing a reference, not passing by reference). Only now realizing my mistake as I read this on a bigger screen.
As to your specific question, the ECMAScript spec is a bit too hard to read in details for me to prove you that you only ever hold a reference to an object value (or anything equivalent: e.g. in Go, a map or a slice is passed by value, but it itself holds a reference to its data, so all those map values you pass by copy everywhere allow you to access and modify the same data, therefore sharing the data between them, as if you had a hold on a reference to the map directly; implementation details are difference, but the big picture is equivalent)
I have a feeling that there is the same root confusion in both cases.
That is that people are introducing references as a way to explain things in terms of another language, say, C++, rather than understanding the language in its own terms.
If it depicts 'passing a reference' not 'passing by reference', the large red text saying "pass by reference" is a bit confusing.
That was the sense of my initial comment: pointing the error in that image.
(remember I though the image "was" the post, and your comment criticizing that there was no "pass by reference" here, to which I replied that I agreed and it was actually "pass a reference")
As far as I am aware, Go also does not have reference values.
Go has pointers, which are references to other values (that can also be pointers). IIRC they are stricter than C++ pointers as you can make a pointer point to an arbitrary address in memory (so in Go it works more as a reference than a pointer to some memory address), and allow pass by reference by way of passing a reference by value (because Go also has address operators that allow you to get a pointer to variable). But Go also have slices where multiple slices can share the same underlying array: how do you can that if not "they all reference the same array"?
Maybe you're reading too much into the word "reference", or maybe others are reading to much into it. It's just a concept after all. What's your precise definition of "reference value" that makes you say that neither JS nor Go have that concept?
In the case of JS, it doesn't really matter whether primitive values are hold by references to them (and thus can possibly be "interned") or copied around, because they're immutable values anyway. It's more important to understand it for composite values like objects, particularly if they're mutable (if they're immutable, it doesn't matter semantically, but as an optimization on memory usage and the time it can take to copy those objects).
Sure there are different reading levels, and many people use those terms/concepts that allow them to port their knowledge of one language (and how it deals with memory) to another, but in the case of this article, it's a bit different: when doing a = b, it's important to understand that both variables share (or reference) the same underlying value (composite, mutable) in memory, so modifying that value through one variable means that the modification can be seen through the other variable.
The post is likely wrong actually for primitive values, depending on the JS runtime implementation: they're most probably not copied, but shared as well, but it doesn't really matter as, as I said, they're immutable anyway.
That is that people are introducing references as a way to explain things in terms of another language, say, C++, rather than understanding the language in its own terms.
People each have their own mental model of how things work, at different levels of abstractions depending on the language and where they come from. What does it mean to "understand the language in its own term" wrt the behavior the OP notices? He explains it with "reference values" vs "primitive values", which one mental model; possibly not accurate depending on how things are actually implemented, but conceptually OK anyway.
My own mental model of this is that there are values that exist "somewhere" in memory, and variables are labels that you put on them (or you could think of yarns, threads, wires attached to them on one end, and to the "code" on the other end; so when entering a function, you put your current state –local variables– aside and hold different yarns in your hands). That works very well for programming languages that abstract away the actual memory allocations, it certainly doesn't work that well for C/C++, but still allows you to understand most C/C++ code anyway.
but in the case of this article, it's a bit different: when doing a = b, it's important to understand that both variables share (or reference) the same underlying value (composite, mutable) in memory
So what this mental model does is to say that the values of a and b aren't the real values of a and b, they're special reference values, and the real, underlying value, is something else which depends on the properties.
Which leads to people making errors like saying that foo(a) is pass by reference because the value you're passing is a 'reference value' and the 'real value' of a is determined by the set of its properties.
But there's nothing in javascript which represents this 'real value', and no operators that operate on this 'real value' (because it doesn't actually exist in the language).
So you end up demoting the actual accessible value and promoting an imaginary and inaccessible kind of value and confusing the language with things like pass-by-copy-of-reference (which seems to be the fallback position when it's pointed out that it isn't actually pass-by-reference).
But at least I'm starting to get an idea of where these peculiar ideas are coming from -- so thanks for being able to talk through it. :)
For further actions, you may consider blocking this person and/or reporting abuse
We're a place where coders share, stay up-to-date and grow their careers.
There is no pass by reference in this article.
Not pass by (no functions in this article) but it's basically the same principle for types and assigning objects to new variables- still just a reference
Not in the least.
What does this print?
Which of your images does it reflect?
An error. That is syntactically wrong
Try now ;)
Well yea, you are kind of comparing apples with pears.
The same thing would happen without the function. For the reference to not break, you cannot reassign
p
- only it's properties... this would still be a reference and mutate a.
Same gist, but without the function
Yeah, when you assigning
{ v: 2 }
to p you're actually changing value of p with completely new one, this diagram can be helpfulWhen you changed value of p with
p = { v: 2 };
it becomes like thismemory location #234 will be inaccessible and garbage collected, I hope this helps
So.
Pass by value, then? :)
Can you take a look in the ecmascript spec and show me where it talks about a being a reference?
I can only see it referring to object values.
Not exactly think it like this
in this code you pass the value of the a and it acts like
addHi("Hello")
so you're not changing the a
In this code it acts like you passed the memory location
foo(#234)
when you change it with
p = { v: 2 };
you're creating entire new object{ v: 2 }
this objects has a new memory location like #427, you're changing value #234 with #427so a is also changingEdit: value of a is not changing
a changes when you change property of a in function
When I run
the output is
I don't see a changing.
So, which bit of this isn't just pass-by-value in both cases?
Again, you reassign the variable. That is indeed pass-by-value but only because you reassign the variable, which in any case you shouldn't do because it's semantically confusing.
Have a look at the style guide by AirBnB for example on this topic (github.com/airbnb/javascript#es6-d...)
In other words (not mine):
Your change does not persist outside of the function-scope because you did not mutate the original reference to
a
. You created a "temporary" version of it inside the function, which doesn't exist anymore because of the function scope. If you want the change to persist (pass-by-ref), you have to dereference it (look at my earlier example).Ok, so we can all agree that it is pass by value.
And that there is no pass by reference happening anywhere here.
Which means that what we're passing is the value of the object.
All well and good.
Can you show me where it says that object values are references, in the ecmascript standard?
I'm having trouble finding support for that claim.
JS always passes function args by value.
But, the "value" of a variable for a JS object is a reference to the underlying object.
This is the same with strings, but they're immutable so their references cause less confusion
So, functions pass by value, but in many cases that value is just a reference to some common structure.
I'm glad we can all agree that javascript doesn't support pass by reference.
Now, could you show me where in the ecmascript standard it talks about an object value being a reference?
Hey man, just some feedback on your conversational style:
I find it's quite confrontational and irritating to read.
To me it feels like your trying to score internet points for being right and aren't actually interested in explaining things to people.
It reads more like a Twitter or Reddit thread and I was always very happy to see a calmer more constructive and friendly style on here.
I really don't mean this as a personal attack, I just feel the conversation could have a lot more merit to others (especially newcomers) if it were more constructive.
Could you be more specific about the confrontational style that you perceive?
A bit of constructive editing...
vvvvvv
Same info and questions, but the second one is showing you're willing to put in some work to find your answer, that we probably fundamentally agree but are stuck on terminology, and will ultimately avoid nag replies like this.
If the language doesn't come naturally to you, try your best to fake it til you make it.
I had a squiz at the spec, couldn't find a specific mention, and ran out of motivation to dig further. The proof though is
For a time, the values of
a
andb
refer to the same object. They hold two pointer-like structures that reference the same shared underlying object. In programming this is often called a reference.I'm signing off the thread btw
That's fine, but for future reference, let's note that a.foo doesn't mean that a is a reference.
It does mean that a and b have the same value, and it does mean that that value is used to find a particular property named 'foo' and modify that.
Which makes me think that all of this 'reference' stuff is something someone made up to try to explain javascript in terms of a language with references, like C++, but which doesn't quite fit.
+1 it's passing a reference by value.
Could you show me where in the ecmascript standard it talks about an object value being a reference?
Wow, for some reason, reading on mobile, I totally missed the article, mixing it up with @alexanderjanke 's comment! (so, my comment was actually discussing the GIF, which depicts passing a reference, not passing by reference). Only now realizing my mistake as I read this on a bigger screen.
As to your specific question, the ECMAScript spec is a bit too hard to read in details for me to prove you that you only ever hold a reference to an object value (or anything equivalent: e.g. in Go, a
map
or a slice is passed by value, but it itself holds a reference to its data, so all thosemap
values you pass by copy everywhere allow you to access and modify the same data, therefore sharing the data between them, as if you had a hold on a reference to the map directly; implementation details are difference, but the big picture is equivalent)If it depicts 'passing a reference' not 'passing by reference', the large red text saying "pass by reference" is a bit confusing.
As far as I am aware, Go also does not have reference values.
Let me know if you can find mention of one in golang.org/ref/spec
I have a feeling that there is the same root confusion in both cases.
That is that people are introducing references as a way to explain things in terms of another language, say, C++, rather than understanding the language in its own terms.
That was the sense of my initial comment: pointing the error in that image.
(remember I though the image "was" the post, and your comment criticizing that there was no "pass by reference" here, to which I replied that I agreed and it was actually "pass a reference")
Go has pointers, which are references to other values (that can also be pointers). IIRC they are stricter than C++ pointers as you can make a pointer point to an arbitrary address in memory (so in Go it works more as a reference than a pointer to some memory address), and allow pass by reference by way of passing a reference by value (because Go also has address operators that allow you to get a pointer to variable). But Go also have slices where multiple slices can share the same underlying array: how do you can that if not "they all reference the same array"?
Maybe you're reading too much into the word "reference", or maybe others are reading to much into it. It's just a concept after all. What's your precise definition of "reference value" that makes you say that neither JS nor Go have that concept?
In the case of JS, it doesn't really matter whether primitive values are hold by references to them (and thus can possibly be "interned") or copied around, because they're immutable values anyway. It's more important to understand it for composite values like objects, particularly if they're mutable (if they're immutable, it doesn't matter semantically, but as an optimization on memory usage and the time it can take to copy those objects).
Sure there are different reading levels, and many people use those terms/concepts that allow them to port their knowledge of one language (and how it deals with memory) to another, but in the case of this article, it's a bit different: when doing
a = b
, it's important to understand that both variables share (or reference) the same underlying value (composite, mutable) in memory, so modifying that value through one variable means that the modification can be seen through the other variable.The post is likely wrong actually for primitive values, depending on the JS runtime implementation: they're most probably not copied, but shared as well, but it doesn't really matter as, as I said, they're immutable anyway.
People each have their own mental model of how things work, at different levels of abstractions depending on the language and where they come from. What does it mean to "understand the language in its own term" wrt the behavior the OP notices? He explains it with "reference values" vs "primitive values", which one mental model; possibly not accurate depending on how things are actually implemented, but conceptually OK anyway.
My own mental model of this is that there are values that exist "somewhere" in memory, and variables are labels that you put on them (or you could think of yarns, threads, wires attached to them on one end, and to the "code" on the other end; so when entering a function, you put your current state –local variables– aside and hold different yarns in your hands). That works very well for programming languages that abstract away the actual memory allocations, it certainly doesn't work that well for C/C++, but still allows you to understand most C/C++ code anyway.
So what this mental model does is to say that the values of a and b aren't the real values of a and b, they're special reference values, and the real, underlying value, is something else which depends on the properties.
Which leads to people making errors like saying that foo(a) is pass by reference because the value you're passing is a 'reference value' and the 'real value' of a is determined by the set of its properties.
But there's nothing in javascript which represents this 'real value', and no operators that operate on this 'real value' (because it doesn't actually exist in the language).
So you end up demoting the actual accessible value and promoting an imaginary and inaccessible kind of value and confusing the language with things like pass-by-copy-of-reference (which seems to be the fallback position when it's pointed out that it isn't actually pass-by-reference).
But at least I'm starting to get an idea of where these peculiar ideas are coming from -- so thanks for being able to talk through it. :)