Specifications for functions: -spec. Efficiently usage - erlang

How could I use -spec word in erlang?
Please give me an idea of efficient usage of this word. Does is stands for documentation purposes only?
I'm try to apply a constraint to function in module by function type specification using -spec, but I've failed - no restrictions have been applied.

-spec attributes are indeed treated by the compiler and the runtime system as documentation. You cannot add any "executable features" to your code using them and the same applies for -type and -opaque attributes.
However they are useful as:
Documentation: they used by EDoc to generate all different forms of documentation for your code. -spec attributes are function signatures which, depending on how much effort you put into them, can make your code more understandable and maintainable. Suppose that your favorite data structure this month is dict(). Consider the following code:
my_function(SomeArg, SomeOtherArg, Dict) ->
...
dict:find(SomeKey, Dict)
...
The variable that is being used as a dict has been named as such. But let's say that you have the following snippet:
my_other_function(NamesDict, PlacesDict) ->
...
R1 = my_function(A, B, NamesDict),
...
R2 = my_function(C, D, PlacesDict),
...
Trying to keep up with this might soon lead to code that repeats this Dict suffix. Even more, you might not even want to remember in the context of my_other_function that the two arguments are dict(). So instead you might want to do this:
-spec my_other_function(dict(), dict()) -> atom().
my_other_function(Names, Places) ->
...
R1 = my_function(A, B, Names),
...
R2 = my_function(C, D, Places),
...
Now it is clear that these arguments should be dict() for the function to work and hopefully everyone will be able to figure that without going deep into the code. But suppose you are using this Name dict() in other places and it stores some particular information that is exposed with different APIs. Then it's a perfect candidate for a -type declaration:
-type names() :: dict().
-spec my_other_function(names(), places()) -> atom().
my_other_function(Names, Places) ->
...
R1 = my_function(A, B, Names),
...
R2 = my_function(C, D, Places),
...
If somebody else makes frequent use of this particular data structure you may want to export it too:
-module(my_module).
-export_type([names/0]).
-type names() :: dict().
Other modules can now refer to this particular data structure:
-module(my_other_module).
-record(my_state, {names :: my_module:names(),
...}).
Finally if you would prefer other developer to not inspect this data structure in any way in their modules, you can declare it as -opaque. Again, this is a "friendly suggestion", as is all the rest of the stuff so far. Or is it...?
Discrepancy detection: If you take time to use -specs and -types you would very much like that these are kept up to date. It is common knowledge that nobody maintains the documentation up to date if there is none watching! Luckily, Dialyzer is watching. Dialyzer can check that in all calls to my_function() the arguments are dict() (it can do this even without your -spec annotations but it's so easier if there are these there too) and scream bloody murder if you call it with something else. It can moreover keep track of these exported types and even report opacity violations. So it's not "just documentation".
Testcase generation: PropEr can use the -spec and -type definitions to automatically check your functions with random testcases. It is capable to make random testcases even from declarations like this one:
-type int_tree() :: {node, integer(), tree(), tree()} | nil.
The brand new way to specify a set of callbacks for a behaviour is by using the familiar -spec syntax. Compiler, Dialyzer and possibly other tools can use this information to check a behaviours implementation. See more in the OTP behaviours code and here
Read more here.

-spec's for functions are specifications which has several places where they help:
They act as documentation of the function. Generating EDoc will pull the specs and make them available in the documentation.
They are a specification for the dialyzer. When the dialyzer runs it will use the specs to determine if the code is wrong in any way. That is, if you spec is wrong - and in some cases it will help the system to understand exactly why the code is wrong too.
They are a valuable tool in the specification of behaviours. There is a new -callback keyword which can be used to do this for behavioural APIs.
They are valuable for constructing a type skeleton of how the program fits together and from where data comes from.
Together with the cousins -type and -opaque you can force certain types to be opaque to pieces of code. That means you are not allowed to see the internal representation on a static verification level. This can in turn help drive modularized code as you are not allowed to tightly couple code pieces.

Related

Partially deferred computation builder

I'm trying to work out how to use a computation builder to represent a deferred, nested set of steps.
I've got the following so far:
type Entry =
| Leaf of string * (unit -> unit)
| Node of string * Entry list * (unit -> unit)
type StepBuilder(desc:string) =
member this.Zero() = Leaf(desc,id)
member this.Bind(v:string, f:unit->string) =
Node(f(), [Leaf(v,id)], id)
member this.Bind(v:Entry, f:unit->Entry) =
match f() with
| Node(label,children,a) -> Node(label, v :: children, a)
| Leaf(label,a) -> Node(label, [v], a)
let step desc = StepBuilder(desc)
let a = step "a" {
do! step "b" {
do! step "c" {
do! step "c.1" {
// todo: this still evals as it goes; need to find a way to defer
// the inner contents...
printfn "TEST"
}
}
}
do! step "d" {
printfn "d"
}
}
This produces the desired structure:
A(B(C(c.1)), D)
My issue is that in building the structure up, the printfn calls are made.
Ideally what I want is to be able to retrieve the tree structure, but be able to call some returned function/s that will then execute the inner blocks.
I realise this means that if you have two nested steps with some "normal" code between them, it would need to be able to read the step declarations, and then invoke over it (if that makes sense?).
I know that Delay and Run are things that are used in deferred execution for computation expressions, but I'm not sure if they help me here, as they unfortunately evaluate for everything.
I'm most likely missing something glaringly obvious and very "functional-y" but I just can't seem to make it do what I want.
Clarification
I'm using id for demonstration, they're part of the puzzle, and I imagine how I might surface the "invokable" parts of my expression that I want.
As mentioned in the other answer, free monads provide a useful theoretical framework for thinking about this kind of problems - however, I think you do not necessarily need them to get an answer to the specific question you are asking here.
First, I had to add Return to your computation builder to make your code compile. As you never return anything, I just added an overload taking unit which is equivalent to Zero:
member this.Return( () ) = this.Zero()
Now, to answer your question - I think you need to modify your discriminated union to allow delaying of computations that produce Entry - you do have functions unit -> unit in the domain model, but that's not quite enough to delay a computation that will produce a new entry. So, I think you need to extend the type:
type Entry =
| Leaf of string * (unit -> unit)
| Node of string * Entry list * (unit -> unit)
| Delayed of (unit -> Entry)
When you are evaluating Entry, you will now need to handle the Delayed case - which contains a function that might perform side-effect such as printing "TEST".
Now you can add Delay to your computation builder and also implement the missing case for Delayed in Bind like this:
member this.Delay(f) = Delayed(f)
member this.Bind(v:Entry, f:unit->Entry) = Delayed(fun () ->
let rec loop = function
| Delayed f -> loop (f())
| Node(label,children,a) -> Node(label, v :: children, a)
| Leaf(label,a) -> Node(label, [v], a)
loop (f()) )
Essentially, Bind will create a new delayed computation that, when called, evaluates the entry v until it finds a node or a leaf (collapsing all other delayed nodes) and then does the same thing as what your code did before.
I think this answers your question - but I'd be a bit careful here. I think computation expressions are useful as a syntactic sugar, but they are very harmful if you think about them more than you think about the domain of the problem that you are actually solving - in the question, you did not say much about your actual problem. If you did, the answer might be very different.
You wrote:
Ideally what I want is to be able to retrieve the tree structure, but be able to call some returned function/s that will then execute the inner blocks.
This is an almost perfect description of the "free monad", which is basically the functional-programming equivalent of the OOP "interpreter pattern". The basic idea behind the free monad is that you convert imperative-style code into a two-step process. The first step builds up an AST, and the second step executes the AST. That way you can do things in between step 1 and step 2, like analyze the tree structure without executing the code. Then when you're ready, you can run your "execute" function, which takes the AST as input and actually does the steps it represents.
I'm not experienced enough with free monads to be able to write a complete tutorial on them, nor to directly answer your question with a step-by-step specific free-monad solution. But I can point you to a few resources that may help you understand the concepts behind them. First, the required Scott Wlaschin link:
https://fsharpforfunandprofit.com/posts/13-ways-of-looking-at-a-turtle-2/#way13
This is the last part of his "13 ways of looking at a turtle" series, where he builds a small LOGO-like turtle-graphics app using many different design styles. In #13, he uses the free-monad style, building it from the ground up so you can see the design decisions that go into that style.
Second, a set of links to Mark Seemann's blog. For the past month or two, Mark Seemann has been writing posts about the free-monad style, though I didn't realize that that's what he was writing about until he was several articles in. There's a terminology difference that may confuse you at first: Scott Wlaschin uses the terms "Stop" and "KeepGoing" for the two possible AST cases ("this is the end of the command list" vs. "there are more commands after this one"). But the traditional names for those two free-monad cases are "Pure" and "Free". IMHO, the names "Pure" and "Free" are too abstract, and I like Scott Wlaschin's "Stop" and "KeepGoing" names better. But I mention this so that when you see "Pure" and "Free" in Mark Seemann's posts, you'll know that it's the same concept as Scott Wlaschin's turtle example.
Okay, with that explanation finished, here are the links to Mark Seemann's posts:
http://blog.ploeh.dk/2017/06/27/pure-times/
http://blog.ploeh.dk/2017/06/28/pure-times-in-haskell/
http://blog.ploeh.dk/2017/07/04/pure-times-in-f/
http://blog.ploeh.dk/2017/07/10/pure-interactions/
http://blog.ploeh.dk/2017/07/11/hello-pure-command-line-interaction/
http://blog.ploeh.dk/2017/07/17/a-pure-command-line-wizard/
http://blog.ploeh.dk/2017/07/24/combining-free-monads-in-haskell/
http://blog.ploeh.dk/2017/07/31/combining-free-monads-in-f/
http://blog.ploeh.dk/2017/08/07/f-free-monad-recipe/
Mark intersperses Haskell examples with F# examples, as you can tell from the URLs. If you are completely unfamiliar with Haskell, you can probably skip those posts as they might confuse you more than they help. But if you've got a passing familiarity with Haskell syntax, seeing the same ideas expressed in both Haskell and F# may help you grasp the concepts better, so I've included the Haskell posts as well as the F# posts.
As I said, I'm not quite familiar enough with free monads to be able to give you a specific answer to your question. But hopefully these links will give you some background knowledge that can help you implement what you're looking for.

Erlang: Will adding type spec to code make dialyzer more effective?

I have a project that doesn't have -spec or -type in code, currently dialyzer can find some warnings, most of them are in machine generated codes.
Will adding type specs to code make dialyzer find more errors?
Off the topic, is there any tool to check whether the specs been violated or not?
Adding typespecs will dramatically improve the accuracy of Dialyzer.
Because Erlang is a dynamic language Dialyzer must default to a rather broad interpretation of types unless you give it hints to narrow the "success" typing that it will go by. Think of it like giving Dialyzer a filter by which it can transform a set of possible successes to a subset of explicit types that should ever work.
This is not the same as Haskell, where the default assumption is failure and all code must be written with successful typing to be compiled at all -- Dialyzer must default to assume success unless it knows for sure that a type is going to fail.
Typespecs are the main part of this, but Dialyzer also checks guards, so a function like
increment(A) -> A + 1.
Is not the same as
increment(A) when A > 100 -> A + 1.
Though both may be typed as
-spec increment(integer()) -> integer().
Most of the time you only care about integer values being integer(), pos_integer(), neg_integer(), or non_neg_integer(), but occasionally you need an arbitrary range bounded only on one side -- and the type language has no way to represent this currently (though personally I would like to see a declaration of 100..infinity work as expected).
The unbounded-range of when A > 100 requires a guard, but a bounded range like when A > 100 and A < 201 could be represented in the typespec alone:
-spec increment(101..200) -> pos_integer().
increment(A) -> %stuff.
Guards are fast with the exception of calling length/1 (which you should probably never actually need in a guard), so don't worry with the performance overhead until you actually know and can demonstrate that you have a performance problem that comes from guards. Using guards and typespecs to constrain Dialyzer is extremely useful. It is also very useful as documentation for yourself and especially if you use edoc, as the typespec will be shown there, making APIs less mysterious and easy to toy with at a glance.
There is some interesting literature on the use of Dialyzer in existing codebases. A well-documented experience is here: Gradual Typing of Erlang Programs: A Wrangler Experience. (Unfortunately some of the other links I learned a lot from previously have disappeared or moved. (!.!) A careful read of the Wrangler paper, skimming over the User's Guide and man page, playing with Dialyzer, and some prior experience in a type system like Haskell's will more than prepare you for getting a lot of mileage out of Dialyzer, though.)
[On a side note, I've spoken with a few folks before about specifying "pure" functions that could be guaranteed as strongly typed either with a notation or by using a different definition syntax (maybe Prolog's :- instead of Erlang's ->... or something), but though that would be cool, and it is very possible even now to concentrate side-effects in a tiny part of the program and pass all results back in a tuple of {Results, SideEffectsTODO}, this is simply not a pressing need and Erlang works pretty darn well as-is. But Dialyzer is indeed very helpful for showing you where you've lost track of yourself!]

Save Mathematica code in `FullForm` syntax

I need to do some metaprogramming on a large Mathematica code base (hundreds of thousands of lines of code) and don't want to have to write a full-blown parser so I was wondering how best to get the code from a Mathematica notebook out in an easily-parsed syntax.
Is it possible to export a Mathematica notebook in FullForm syntax, or to save all definitions in FullForm syntax?
The documentation for Save says that it can only export in the InputForm syntax, which is non-trivial to parse.
The best solution I have so far is to evaluate the notebook and then use DownValues to extract the rewrite rules with arguments (but this misses symbol definitions) as follows:
DVs[_] := {}
DVs[s_Symbol] := DownValues[s]
stream = OpenWrite["FullForm.m"];
WriteString[stream,
DVs[Symbol[#]] & /# Names["Global`*"] // Flatten // FullForm];
Close[stream];
I've tried a variety of approaches so far but none are working well. Metaprogramming in Mathematica seems to be extremely difficult because it keeps evaluating things that I want to keep unevaluated. For example, I wanted to get the string name of the infinity symbol using SymbolName[Infinity] but the Infinity gets evaluated into a non-symbol and the call to SymbolName dies with an error. Hence my desire to do the metaprogramming in a more suitable language.
EDIT
The best solution seems to be to save the notebooks as package (.m) files by hand and then translate them using the following code:
stream = OpenWrite["EverythingFullForm.m"];
WriteString[stream, Import["Everything.m", "HeldExpressions"] // FullForm];
Close[stream];
You can certainly do this. Here is one way:
exportCode[fname_String] :=
Function[code,
Export[fname, ToString#HoldForm#FullForm#code, "String"],
HoldAllComplete]
For example:
fn = exportCode["C:\\Temp\\mmacode.m"];
fn[
Clear[getWordsIndices];
getWordsIndices[sym_, words : {__String}] :=
Developer`ToPackedArray[words /. sym["Direct"]];
];
And importing this as a string:
In[623]:= Import["C:\\Temp\\mmacode.m","String"]//InputForm
Out[623]//InputForm=
"CompoundExpression[Clear[getWordsIndices], SetDelayed[getWordsIndices[Pattern[sym, Blank[]], \
Pattern[words, List[BlankSequence[String]]]], Developer`ToPackedArray[ReplaceAll[words, \
sym[\"Direct\"]]]], Null]"
However, going to other language to do metaprogramming for Mathematica sounds ridiculous to me, given that Mathematica is very well suited for that. There are many techniques available in Mathematica to do meta-programming and avoid premature evaluation. One that comes to my mind I described in this answer, but there are many others. Since you can operate on parsed code and use the pattern-matching in Mathematica, you save a lot. You can browse the SO Mathematica tags (past questions) and find lots of examples of meta-programming and evaluation control.
EDIT
To ease your pain with auto-evaluating symbols (there are only a few actually, Infinity being one of them).If you just need to get a symbol name for a given symbol, then this function will help:
unevaluatedSymbolName = Function[sym, SymbolName#Unevaluated#sym, HoldAllComplete]
You use it as
In[638]:= unevaluatedSymbolName[Infinity]//InputForm
Out[638]//InputForm="Infinity"
Alternatively, you can simply add HoldFirst attribute to SymbolName function via SetAttributes. One way is to do that globally:
SetAttributes[SymbolName,HoldFirst];
SymbolName[Infinity]//InputForm
Modifying built-in functions globally is however dangerous since it may have unpredictable effects for such a large system as Mathematica:
ClearAttributes[SymbolName, HoldFirst];
Here is a macro to use that locally:
ClearAll[withUnevaluatedSymbolName];
SetAttributes[withUnevaluatedSymbolName, HoldFirst];
withUnevaluatedSymbolName[code_] :=
Internal`InheritedBlock[{SymbolName},
SetAttributes[SymbolName, HoldFirst];
code]
Now,
In[649]:=
withUnevaluatedSymbolName[
{#,StringLength[#]}&[SymbolName[Infinity]]]//InputForm
Out[649]//InputForm= {"Infinity", 8}
You may also wish to do some replacements in a piece of code, say, replace a given symbol by its name. Here is an example code (which I wrap in Hold to prevent it from evaluation):
c = Hold[Integrate[Exp[-x^2], {x, -Infinity, Infinity}]]
The general way to do replacements in such cases is using Hold-attributes (see this answer) and replacements inside held expressions (see this question). For the case at hand:
In[652]:=
withUnevaluatedSymbolName[
c/.HoldPattern[Infinity]:>RuleCondition[SymbolName[Infinity],True]
]//InputForm
Out[652]//InputForm=
Hold[Integrate[Exp[-x^2], {x, -"Infinity", "Infinity"}]]
, although this is not the only way to do this. Instead of using the above macro, we can also encode the modification to SymbolName into the rule itself (here I am using a more wordy form ( Trott - Strzebonski trick) of in-place evaluation, but you can use RuleCondition as well:
ClearAll[replaceSymbolUnevaluatedRule];
SetAttributes[replaceSymbolUnevaluatedRule, HoldFirst];
replaceSymbolUnevaluatedRule[sym_Symbol] :=
HoldPattern[sym] :> With[{eval = SymbolName#Unevaluated#sym}, eval /; True];
Now, for example:
In[629]:=
Hold[Integrate[Exp[-x^2],{x,-Infinity,Infinity}]]/.
replaceSymbolUnevaluatedRule[Infinity]//InputForm
Out[629]//InputForm=
Hold[Integrate[Exp[-x^2], {x, -"Infinity", "Infinity"}]]
Actually, this entire answer is a good demonstration of various meta-programming techniques. From my own experiences, I can direct you to this, this, this, this and this answers of mine, where meta-programming was essential to solve problem I was addressing. You can also judge by the fraction of functions in Mathematica carrying Hold-attributes to all functions - it is about 10-15 percents if memory serves me well. All those functions are effectively macros, operating on code. To me, this is a very indicative fact, telling me that Mathematica jeavily builds on its meta-programming facilities.
The full forms of expressions can be extracted from the Code and Input cells of a notebook as follows:
$exprs =
Cases[
Import["mynotebook.nb", "Notebook"]
, Cell[content_, "Code"|"Input", ___] :>
ToExpression[content, StandardForm, HoldComplete]
, Infinity
] //
Flatten[HoldComplete ## #, 1, HoldComplete] & //
FullForm
$exprs is assigned the expressions read, wrapped in Hold to prevent evaluation. $exprs could then be saved into a text file:
Export["myfile.txt", ToString[$exprs]]
Package files (.m) are slightly easier to read in this way:
Import["mypackage.m", "HeldExpressions"] //
Flatten[HoldComplete ## #, 1, HoldComplete] &

What alternatives are there to parameterised modules in Erlang?

I can see why parameterised modules are used so much, as they allow us to do things like:
X = y:new("param").
X:action1().
X.get_property():
: which feels very OO. However, this is only an experimental feature in Erlang and I hear it may be removed, so I need to find an alternative.
Parametrised modules are nothing but a shortcut on the first argument of a function. See these two examples:
-module(y, [Name,Age]).
-compile(export_all).
action1() -> io:format("Hello, ~p!~n",[Name]).
get_age() -> Age.
Run it:
1> c(y).
{ok,y}
2> X = y:new("Fred",22).
{y,"Fred",22}
3> X:action1().
Hello, "Fred"!
ok
4> X:get_age().
22
Now without parametrized stuff:
-module(y).
-compile(export_all).
action1([Name,_]) -> io:format("Hello, ~p!~n",[Name]).
get_age([_,Age]) -> Age.
Running it:
1> c(y).
{ok,y}
2> X = ["Fred",22].
["Fred",22]
3> y:action1(X).
Hello, "Fred"!
ok
4> y:get_age(X).
22
The biggest 'advantage' of parametrized modules is that you shift the burden of carrying state from a variable to a module name. This is in appearance much simpler to someone not used to 'the Erlang way', but it clashes with conventional styles of code.
It's not just a question of being experimental or not. You throw out referential transparency and the semantics to immutable variables become a bit weird. A good example of this is to imagine you add the following functions to a parametrized module:
ret_fn() -> fun(Age) -> Age + 5 end.
When compiling the module, you get the warning ./y.erl:8: Warning: variable 'Age' shadowed in 'fun'. This is warning you that you are using the name of a predefined variable within the head clause of an anonymous function. However, a quick look at the ret_fn/0 function shows absolutely NO sign of where that variable is from.
Now imagine that you use the variable Name for any other purpose; you will get a run time error telling you ** error: no match of right hand side value < ... >.
The point I'm making is that parametrized modules reduce the amount of typing you need to do at the expense of logical simplicity. Not just for you, but for any other Erlang programmer to work on your code.
On top of that, tools like dialyzer, TypEr, tidiers and whatnot have no guarantee to support these idioms. These tools are pretty useful, too! Don't dismiss them. (Edit: newer versions of Erlang (R13B04+) now guarantee that support)
The best alternative to parametrized modules is to avoid them and use what every other Erlang programmer outside of mochiweb is using.
Why can't you use the usual way i.e. send message to a (registered) server process associated with a module?
The message in question can be whatever you want of course e.g. configuration etc.
What problem are you trying to solve that can't be handled in such way?

Hidden Features of F#

Locked. This question and its answers are locked because the question is off-topic but has historical significance. It is not currently accepting new answers or interactions.
This is the unabashed attempt of a similar C# question.
So what are your favorite F# hidden (or not) features?
Most of the features I've used so far aren't exactly hidden but have been quite refreshing. Like how trivial it is to overload operators compared to say C# or VB.NET.
And Async<T> has helped me shave off some real ugly code.
I'm quite new to the language still so it'd be great to learn what other features are being used in the wild.
User defined numeric literals can be defined by providing a module whose name starts with NumericLiteral and which defines certain methods (FromZero, FromOne, etc.).
In particular, you can use this to provide a much more readable syntax for calling LanguagePrimitives.GenericZero and LanguagePrimitives.GenericOne:
module NumericLiteralG = begin
let inline FromZero() = LanguagePrimitives.GenericZero
let inline FromOne() = LanguagePrimitives.GenericOne
end
let inline genericFactorial n =
let rec fact n = if (n = 0G) then 1G else n * (fact (n - 1G))
fact n
let flt = genericFactorial 30.
let bigI = genericFactorial 30I
F# has a little-used feature called "signature files". You can have a big implementation file full of public types/methods/modules/functions, but then you can hide and selectively expose that functionality to the sequel of the program via a signature file. That is, a signature file acts as a kind of screen/filter that enables you to make entities "public to this file" but "private to the rest of the program".
I feel like this is a pretty killer feature on the .Net platform, because the only other/prior tool you have for this kind of encapsulation is assemblies. If you have a small component with a few related types that want to be able to see each other's internal details, but don't want those types to have all those bits public to everyone, what can you do? Well, you can do two things:
You can put that component in a separate assembly, and make the members that those types share be "internal", and make the narrow part you want everyone else to see be "public", or
You just mark the internal stuff "internal" but you leave those types in your gigantic assembly and just hope that all the other code in the assembly chooses not to call those members that were only marked 'internal' because one other type needed to see it.
In my experience, on large software projects, everyone always does #2, because #1 is a non-starter for various reasons (people don't want 50 small assemblies, they want 1 or 2 or 3 large assemblies, for other maybe-good reasons unrelated to the encapsulation point I am raising (aside: everyone mentions ILMerge but no one uses it)).
So you chose option #2. Then a year later, you finally decide to refactor out that component, and you discover that over the past year, 17 other places now call into that 'internal' method that was really only meant for that one other type to call, making it really hard to factor out that bit because now everyone depends on those implementation details. Bummer.
The point is, there is no good way to create a moderate-size intra-assembly encapsulation scope/boundary in .Net. Often times "internal" is too big and "private" is too small.
... until F#. With F# signature files, you can create an encapsulation scope of "this source code file" by marking a bunch of stuff as public within the implementation file, so all the other code in the file can see it and party on it, but then use a signature file to hide all of the details expect the narrow public interface that component exposes to the rest of the world. This is happy. Define three highly related types in one file, let them see each others implementation details, but only expose the truly public stuff to everyone else. Win!
Signature files are perhaps not the ideal feature for intra-assembly encapsulation boundaries, but they are the only such feature I know, and so I cling to them like a life raft in the ocean.
TL;DR
Complexity is the enemy. Encapsulation boundaries are a weapon against this enemy. "private" is a great weapon but sometimes too small to be applicable, and "internal" is often too weak because so much code (entire assembly and all InternalsVisibleTo's) can see internal stuff. F# offers a scope bigger than "private to a type" but smaller than "the whole assembly", and that is very useful.
I wonder what happens if you add
<appSettings>
<add key="fsharp-navigationbar-enabled" value="true" />
</appSettings>
to your devenv.exe.config file? (Use at your own risk.)
Passing --warnon:1182 to the compiler turns on warnings about unused variables; variable names that begin with underscore are immune.
Automatically-generated comparison functions for algebraic data types (based on lexicographical ordering) is a nice feature that is relatively unknown; see
http://lorgonblog.spaces.live.com/blog/cns!701679AD17B6D310!548.entry
for an example.
Yes, F# doesn't have any 'hidden' features, but it sure does have a lot of power packed into the simple language. A less-known feature of the language, is where you can basically enable duck typing despite the fact F# is staticaly typed.
See this question
F# operator "?"
for info on the question-mark operator and how it provides the basic language mechanism to build a feature akin to 'dynamic' in C#.
Not really hidden, but as a non-ML person this escaped me for quite a while:
Pattern matching can decompose arbitrarily deep into data structures.
Here's a [incredibly arbitrary] nested tuple example; this works on lists or unions or any combinations of nested values:
let listEven =
"Manipulating strings can be intriguing using F#".Split ' '
|> List.ofArray
|> List.map (fun x -> (x.Length % 2 = 0, x.Contains "i"), x)
|> List.choose
( function (true, true), s -> Some s
| _, "F#" -> Some "language"
| _ -> None )
Use of F# as a utility scripting language may be under appreciated. F# enthusiasts tend to be quants. Sometimes you want something to back up your MP3s (or dozens of database servers) that's a little more robust than batch. I've been hunting for a modern replacement for jscript / vbscript. Lately, I've used IronPython, but F# may be more complete and the .NET interaction is less cumbersome.
I like curried functions for entertainment value. Show a curried function to a pure procedural / OOP program for at least three WTFs. Starting with this is a bad way to get F# converts, though :)
Inlined operators on generic types can have different generic constraints:
type 'a Wrapper = Wrapper of 'a with
static member inline (+)(Wrapper(a),Wrapper(b)) = Wrapper(a + b)
static member inline Exp(Wrapper(a)) = Wrapper(exp a)
let objWrapper = Wrapper(obj())
let intWrapper = (Wrapper 1) + (Wrapper 2)
let fltWrapper = exp (Wrapper 1.0)
(* won''t compile *)
let _ = exp (Wrapper 1)
There are no hidden features, because F# is in design mode. All what we have is a Technical Preview, which changes every two month.
see http://research.microsoft.com/fsharp/

Resources