This question has been asked for other languages. I would like to ask this in relation to Julia.
What are the general guidelines for choosing between an array of struct e.g.
struct vertex
x::Real
y::Real
gradient_x::Real
gradient_y::Real
end
myarray::Array{Vertex}
and multiple arrays.
xpositions::Array{<:Real}
ypositions::Array{<:Real}
gradient_x::Array{<:Real}
gradient_y::Array{<:Real}
Are there any performance considerations? Or is it just a style/readability issue.
Your struct as it currently stands will perform poorly. From the Performance Tips you should always:
Avoid fields with abstract type
Similarly, you should always prefer Vector{<:Real} to Vector{Real}.
The Julian way to approach this is to parameterize your struct as follows:
struct Vertex{T<:Real}
x::T
y::T
gradient_x::T
gradient_y::T
end
Given the above, the two approaches discussed in the question will now have roughly similar performance. In practice, it really depends on what kind of operations you want to perform. For example, if you frequently need a vector of just the x fields, then having multiple arrays will probably be a better approach, since any time you need a vector of x fields from a Vector{Vertex} you will need to loop over the structs to allocate it:
xvec = [ v.x for v in vertexvec ]
On the other hand, if your application lends itself to functions called over all four fields of the struct, then your code will be significantly cleaner if you use a Vector{Vertex} and will be just as performant as looping over 4 arrays. Broadcasting in particular will make for nice clean code here. For example, if you have some function:
f(x, y, gradient_x, gradient_y)
then just add the method:
f(v::Vertex) = f(v.x, v.y, v.gradient_x, v.gradient_y)
Now if you want to apply it to vv::Vector{Vertex}, you can just use:
f.(vv)
Remember, user-defined types in Julia are just as performant as "in-built" types. In fact, many types that you might think of as in-built are just defined in Julia itself, much as you are doing here.
So the short summary is: both approaches are performant, so use whichever makes more sense in the context of your application.
Related
As I'm learning Julia, I am wondering how to properly do things I might have done in Python, Java or C++ before. For example, previously I might have used an abstract base class (or interface) to define a family of models through classes. Each class might then have a method like calculate. So to call it I might have model.calculate(), where the model is an object from one of the inheriting classes.
I get that Julia uses multiple dispatch to overload functions with different signatures such as calculate(model). The question I have is how to create different models. Do I use the type system for that and create different types like:
abstract type Model end
type BlackScholes <: Model end
type Heston <: Model end
where BlackScholes and Heston are different types of model? If so, then I can overload different calculate methods:
function calculate(model::BlackScholes)
# code
end
function calculate(model::Heston)
# code
end
But I'm not sure if this is a proper and idiomatic use of types in Julia. I will greatly appreciate your guidance!
This is a hard question to answer. Julia offers a wide range of tools to solve any given problem, and it would be hard for even a core developer of the language to assert that one particular approach is "right" or even "idiomatic".
For example, in the realm of simulating and solving stochastic differential equations, you could look at the approach taken by Chris Rackauckas (and many others) in the suite of packages under the JuliaDiffEq umbrella. However, many of these people are extremely experienced Julia coders, and what they do may be somewhat out of reach for less experienced Julia coders who just want to model something in a manner that is reasonably sensible and attainable for a mere mortal.
It is is possible that the only "right" answer to this question is to direct users to the Performance Tips section of the docs, and then assert that as long as you aren't violating any of the recommendations there, then what you are doing is probably okay.
I think the best way I can answer this question from my own personal experience is to provide an example of how I (a mere mortal) would approach the problem of simulating different Ito processes. It is actually not too far off what you have put in the question, although with one additional layer. To be clear, I make no claim that this is the "right" way to do things, merely that it is one approach that utilizes multiple dispatch and Julia's type system in a reasonably sensible fashion.
I start off with an abstract type, for nesting specific subtypes that represent specific models.
abstract type ItoProcess ; end
Now I define some specific model subtypes, e.g.
struct GeometricBrownianMotion <: ItoProcess
mu::Float64
sigma::Float64
end
struct Heston <: ItoProcess
mu::Float64
kappa::Float64
theta::Float64
xi::Float64
end
Note, in this case I don't need to add constructors that convert arguments to Float64, since Julia does this automatically, e.g. GeometricBrownianMotion(1, 2.0) will work out-of-the-box, as Julia will automatically convert 1 to 1.0 when constructing the type.
However, I might want to add some constructors for common parameterizations, e.g.
GeometricBrownianMotion() = GeometricBrownianMotion(0.0, 1.0)
I might also want some functions that return useful information about my models, e.g.
number_parameter(model::GeometricBrownianMotion) = 2
number_parameter(model::Heston) = 4
In fact, given how I've defined the models above, I could actually be a bit sneaky and define a method that works for all subtypes:
number_parameter(model::T) where {T<:ItoProcess} = length(fieldnames(typeof(model)))
Now I want to add some code that allows me to simulate my models:
function simulate(model::T, numobs::Int, stval) where {T<:ItoProcess}
# code here that is common to all subtypes of ItoProcess
simulate_inner(model, somethingelse)
# maybe more code that is common to all subtypes of ItoProcess
end
function simulate_inner(model::GeometricBrownianMotion, somethingelse)
# code here that is specific to GeometricBrownianMotion
end
function simulate_inner(model::Heston, somethingelse)
# code here that is specific to Heston
end
Note that I have used the abstract type to allow me to group all code that is common to all subtypes of ItoProcess in the simulate function. I then use multiple dispatch and simulate_inner to run any code that needs to be specific to a particular subtype of ItoProcess. For the aforementioned reasons, I hesitate to use the phrase "idiomatic", but let me instead say that the above is quite a common pattern in typical Julia code.
The one thing to be careful of in the above code is to ensure that the output type of the simulate function is type-stable, that is, the output type can be uniquely determined by the input types. Type stability is usually an important factor in ensuring performant Julia code. An easy way in this case to ensure type-stability is to always return Matrix{Float64} (if the output type is fixed for all subtypes of ItoProcess then obviously it is uniquely determined). I examine a case where the output type depends on input types below for my estimate example. Anyway, for simulate I might always return Matrix{Float64} since for GeometricBrownianMotion I only need one column, but for Heston I will need two (the first for price of the asset, the second for the volatility process).
In fact, depending on how the code is used, type-stability is not always necessary for performant code (see eg using function barriers to prevent type-instability from flowing through to other parts of your program), but it is a good habit to be in (for Julia code).
I might also want routines to estimate these models. Again, I can follow the same approach (but with a small twist):
function estimate(modeltype::Type{T}, data)::T where {T<:ItoProcess}
# again, code common to all subtypes of ItoProcess
estimate_inner(modeltype, data)
# more common code
return T(some stuff generated from function that can be used to construct T)
end
function estimate_inner(modeltype::Type{GeometricBrownianMotion}, data)
# code specific to GeometricBrownianMotion
end
function estimate_inner(modeltype::Type{Heston}, data)
# code specific to Heston
end
There are a few differences from the simulate case. Instead of inputting an instance of GeometricBrownianMotion or Heston, I instead input the type itself. This is because I don't actually need an instance of the type with defined values for the fields. In fact, the values of those fields is the very thing I am attempting to estimate! But I still want to use multiple dispatch, hence the ::Type{T} construct. Note also I have specified an output type for estimate. This output type is dependent on the ::Type{T} input, and so the function is type-stable (output type can be uniquely determined by input types). But common with the simulate case, I have structured the code so that code that is common to all subtypes of ItoProcess only needs to be written once, and code that is specific to the subtypes is separted out.
This answer is turning into an essay, so I should tie it off here. Hopefully this is useful to the OP, as well as anyone else getting into Julia. I just want to finish by emphasizing that what I have done above is only one approach, there are others that will be just as performant, but I have personally found the above to be useful from a structural perspective, as well as reasonably common across the Julia ecosystem.
Suppose A is an abstract type, I have a function f{T<:A}(x::Vector{A}). So x could be type Vector{A} or Vector{B} where B <: A. In the middle of the function I would like to cast x to Vector{A} so it can be consumed by another function that requires that signature.
What's the best way to do that? At the moment I am doing x = collect(A, x). Is there a way to avoid copying if possible?
If at all possible, I'd just change your second function definition to be parametric like f. Enforcing this kind of container structure in method signatures is a big performance bug that doesn't gain you any functionality… and just makes them much harder to use.
That said, the best way to do this kind of conversion where you don't care if the output aliases the input is with convert(Vector{A}, x). This will be a no-op if x already isa Vector{A}, but otherwise it'll be just like collect. That's as good as it gets.
Here's why: two containers of types Vector{A} and Vector{B} cannot share the same memory if A !== B since it'd be possible to corrupt the data in the Vector{B} by assigning a non-B element to the array through the Vector{A}.
I am attempting to represent dice rolls in Julia. I am generating all the rolls of a ndsides with
sort(collect(product(repeated(1:sides, n)...)), by=sum)
This produces something like:
[(1,1),(2,1),(1,2),(3,1),(2,2),(1,3),(4,1),(3,2),(2,3),(1,4) … (6,3),(5,4),(4,5),(3,6),(6,4),(5,5),(4,6),(6,5),(5,6),(6,6)]
I then want to be able to reasonably modify those tuples to represent things like dropping the lowest value in the roll or adding a constant number, etc., e.g., converting (2,5) into (10,2,5) or (5,).
Does Julia provide nice functions to easily modify (not necessarily in-place) n-tuples or will it be simpler to move to a different structure to represent the rolls?
Thanks.
Tuples are immutable, so you can't modify them in-place. There is very good support for other mutable data structures, so there aren't many methods that take a tuple and return a new, slightly modified copy. One way to do this is by splatting a section of the old tuple into a new tuple, so, for example, to create a new tuple like an existing tuple t but with the first element set to 5, you would write: tuple(5, t[2:end]...). But that's awkward, and there are much better solutions.
As spencerlyon2 suggests in his comment, a one dimensional Array{Int,1} is a great place to start. You can take a look at the Data Structures manual page to get an idea of the kinds of operations you can use; one-dimensional Arrays are iterable, indexable, and support the dequeue interface.
Depending upon how important performance is and how much work you're doing, it may be worthwhile to create your own data structure. You'll be able to add your own, specific methods (e.g., reroll!) for that type. And by taking advantage of some of the domain restrictions (e.g., if you only ever want to have a limited number of dice rolls), you may be able to beat the performance of the general Array.
You can construct a new tuple based on spreading or slicing another:
julia> b = (2,5)
(2, 5)
julia> (10, b...)
(10, 2, 5)
julia> b[2:end]
(5,)
Assuming you want a list of arrays, each having the same size. Is it better performance-wise to use a 2D array :
integer, allocatable :: data(:,:)
or an array of derived types :
type test
integer, allocatable :: content(:)
end type
type(test), allocatable :: data(:)
Of course, for arrays of different sizes, we don't have a choice. But how is the memory managed between the 2 cases ? Also, is one of them good code practice ?
Choose the implementation which minimises the conceptual distance that your mind has to leap between the problem in your head and the solution in your code. The force of this approach increases with age, both the age of your code (good conceptual design is a solid foundation for future development) and your own age (the less effort understanding your code demands the longer you'll remain mentally competent enough to understand it).
As to the non-opinion-determined part of your question concerning the way that the memory is managed ... My naive expectation is that most compilers will, under most circumstances, allocate contiguous memory for the first of your outlines, and may not for the second. But I don't care enough about this to check, and I do not think that you should either. I don't, by this, suggest that you should not be interested in what is going on under the hood, but rather that you should be more concerned with the matters referred to in the first paragraph.
In general, you want to use the simplest data structure that suits your problem. If a 2d rectangular array meets your needs - and for a huge number of scientific computing problems, problems for which Fortran is a good choice, it does - then that's the choice you want.
The 2d array will be contiguous in memory, which will normally make accessing it faster both due to caching and one fewer level of indirection; the 2d array will also allow you to do things like data = data * 2 or data = 0. which the array-of-array approach doesn't [Edited to add: though as IanH points out in comments you can create a defined type and defined operations on those types to allow this]. Those advantages are great enough that even when you have "ragged arrays", if the range of expected row lengths isn't that large, implementing it as a rectangular 2d array is sometimes a choice worth considering.
Why is the Haskell implementation so focused on linked lists?
For example, I know Data.Sequence is more efficient
with most of the list operations (except for the cons operation), and is used a lot;
syntactically, though, it is "hardly supported". Haskell has put a lot of effort into functional abstractions, such as the Functor and the Foldable class, but their syntax is not compatible with that of the default list.
If, in a project I want to optimize and replace my lists with sequences - or if I suddenly want support for infinite collections, and replace my sequences with lists - the resulting code changes are abhorrent.
So I guess my wondering can be made concrete in questions such as:
Why isn't the type of map equal to (Functor f) => (a -> b) -> f a -> f b?
Why can't the [] and (:) functions be used for, for example, the type in Data.Sequence?
I am really hoping there is some explanation for this, that doesn't include the words "backwards compatibility" or "it just grew that way", though if you think there isn't, please let me know. Any relevant language extensions are welcome as well.
Before getting into why, here's a summary of the problem and what you can do about it. The constructors [] and (:) are reserved for lists and cannot be redefined. If you plan to use the same code with multiple data types, then define or choose a type class representing the interface you want to support, and use methods from that class.
Here are some generalized functions that work on both lists and sequences. I don't know of a generalization of (:), but you could write your own.
fmap instead of map
mempty instead of []
mappend instead of (++)
If you plan to do a one-off data type replacement, then you can define your own names for things, and redefine them later.
-- For now, use lists
type List a = [a]
nil = []
cons x xs = x : xs
{- Switch to Seq in the future
-- type List a = Seq a
-- nil = empty
-- cons x xs = x <| xs
-}
Note that [] and (:) are constructors: you can also use them for pattern matching. Pattern matching is specific to one type constructor, so you can't extend a pattern to work on a new data type without rewriting the pattern-matchign code.
Why there's so much list-specific stuff in Haskell
Lists are commonly used to represent sequential computations, rather than data. In an imperative language, you might build a Set with a loop that creates elements and inserts them into the set one by one. In Haskell, you do the same thing by creating a list and then passing the list to Set.fromList. Since lists so closely match this abstraction of computation, they have a place that's unlikely to ever be superseded by another data structure.
The fact remains that some functions are list-specific when they could have been generic. Some common functions like map were made list-specific so that new users would have less to learn. In particular, they provide simpler and (it was decided) more understandable error messages. Since it's possible to use generic functions instead, the problem is really just a syntactic inconvenience. It's worth noting that Haskell language implementations have very little list-speficic code, so new data structures and methods can be just as efficient as the "built-in" ones.
There are several classes that are useful generalizations of lists:
Functor supplies fmap, a generalization of map.
Monoid supplies methods useful for collections with list-like structure. The empty list [] is generalized to other containers by mempty, and list concatenation (++) is generalized to other containers by mappend.
Applicative and Monad supply methods that are useful for interpreting collections as computations.
Traversable and Foldable supply useful methods for running computations over collections.
Of these, only Functor and Monad were in the influential Haskell 98 spec, so the others have been overlooked to varying degrees by library writers, depending on when the library was written and how actively it was maintained. The core libraries have been good about supporting new interfaces.
I remember reading somewhere that map is for lists by default since newcomers to Haskell would be put off if they made a mistake and saw a complex error about "Functors", which they have no idea about. Therefore, they have both map and fmap instead of just map.
EDIT: That "somewhere" is the Monad Reader Issue 13, page 20, footnote 3:
3You might ask why we need a separate map function. Why not just do away with the current
list-only map function, and rename fmap to map instead? Well, that’s a good question. The
usual argument is that someone just learning Haskell, when using map incorrectly, would much
rather see an error about lists than about Functors.
For (:), the (<|) function seems to be a replacement. I have no idea about [].
A nitpick, Data.Sequence isn't more efficient for "list operations", it is more efficient for sequence operations. That said, a lot of the functions in Data.List are really sequence operations. The finger tree inside Data.Sequence has to do quite a bit more work for a cons (<|) equivalent to list (:), and its memory representation is also somewhat larger than a list as it is made from two data types a FingerTree and a Deep.
The extra syntax for lists is fine, it hits the sweet spot at what lists are good at - cons (:) and pattern-matching from the left. Whether or not sequences should have extra syntax is further debate, but as you can get a very long way with lists, and lists are inherently simple, having good syntax is a must.
List isn't an ideal representation for Strings - the memory layout is inefficient as each Char is wrapped with a constructor. This is why ByteStrings were introduced. Although they are laid out as an array ByteStrings have to do a bit of administrative work - [Char] can still be competitive if you are using short strings. In GHC there are language extensions to give ByteStrings more String-like syntax.
The other major lazy functional Clean has always represented strings as byte arrays, but its type system made this more practical - I believe the ByteString library uses unsafePerfomIO under the hood.
With version 7.8, ghc supports overloading list literals, compare the manual. For example, given appropriate IsList instances, you can write
['0' .. '9'] :: Set Char
[1 .. 10] :: Vector Int
[("default",0), (k1,v1)] :: Map String Int
['a' .. 'z'] :: Text
(quoted from the documentation).
I am pretty sure this won't be an answer to your question, but still.
I wish Haskell had more liberal function names(mixfix!) a la Agda. Then, the syntax for list constructors (:,[]) wouldn't have been magic; allowing us to at least hide the list type and use the same tokens for our own types.
The amount of code change while migrating between list and custom sequence types would be minimal then.
About map, you are a bit luckier. You can always hide map, and set it equal to fmap yourself.
import Prelude hiding(map)
map :: (Functor f) => (a -> b) -> f a -> f b
map = fmap
Prelude is great, but it isn't the best part of Haskell.