Python,:

No switch keyword. And NO, dictionary is not a replacement for it. Not even a bunch of elif statements. Inconsistent line break handling. Why can I do: test = (1, 2, 3) And not: from itertools import cycle, islice, izip Why can't I do: if stuff \ and foo \ or bar: return "Formated string with %(arg)s" % \ {'arg': "bloody slash"} without using slashes? There is not one obvious and only one way to do it. Python fails on its motto just like Java failed on "Write once run anywhere". # what somebody from an another language would do if not test.has_key('foo'): test['foo'] = 0 n = test['foo'] = test['foo'] + 1 vs # what an agnostic beginer would do try: test['foo'] += 1 except KeyError: test['foo'] = 1 n = test['foo'] vs # what you end up after looking for dictionary default value in the python doc test.setdefault('foo', 0) n = test['foo'] = test['foo'] + 1 vs # what I would do n = test['foo'] = test.get('foo', 0) + 1 And the worst is that they don't do exactly the same thing. There are subtle differences. Choice between spaces and tabs. There should be no choice. Pick on, set it in stone and stop fighting. Why can you do that: test = {} test['foo'] = 0 but not: test = [] test[] = 0

附注:" ".join(l)是优秀的人。不要抱怨它，这并不明显，但考虑到迭代器模式，这是正确的方法。

Erlang不在此列表中。在我最喜欢的语言中，但有一些缺陷是肯定的:

Syntax. This includes the 3 terminating tokens (,;.) and aesthetics, but more generally on how the semantic meaning of the code is expressed in text. An example is on how all lowercase tokens are atoms, so to refer to a function you can't just name it, you have to fun my_function/1, and ?PRECEDE_CONSTANTS_WITH_QUESTION_MARKS. Coming from Scheme, Haskell, etc. you just wish you could use a name. Library support is lame. This is mostly external libraries, but even the old standard library. Newer versions of Erlang have sensible POSIX regexes, but the old one had a pretty horrible library for basic string manipulation. You also never know when you're getting the value, or {ok, Value}. Related: non-uniform tools for building and distribution. Ruby has gem and rake, RSpec. Perl has CPAN. I'm unaware of decent equivalents in Erlang. The few Erlang specific tools are pretty strange. Mnesia is a great database, but coming from SQL you have lots of trivialities to learn. Same with the documentation @spec, which has a strange way of describing signatures. Often the functional paradigm hurts when you just want that little bit of mutation. Supposing you want a Hash Table, you can't just hack it as in Scheme, or SML. ets and dets alleviate some of the pain, but not much.

第六,奖金:

模块的导入和导出语法是一堆失败，这与Java的80多行导入语句没有什么不同。

综上所述，Erlang是一种乐趣^_^

C

No parametric polymorphism (i.e. C++ templates). It makes writing reusable data structures and algorithms a pain (and there's hardly any static checking). See for instance the comparator argument to qsort and bsearch: the comparator takes void pointers :( No library of data structures. I really hate writing my own hash table. I also really hate scouring the web for a library of reusable data structures. Especially if it turns out to be incomplete. Strings. Inefficient representation, unwieldy if you make it sane, too hard to safely input a string. No standard for snprintf. Too hard to create a format string with sprintf, then use that to create a string with sprintf again, in a safe way. Only lexical macros. If different compilers expects function annotation in different places, I have to put the same HAS_NO_SIDE_EFFECTS in different places. Why can't I just grab the function, switch over the compiler type, and then insert it at the right place by a macro call? No portable libraries for common functionality. For sockets and threading, I use SDL---a frigging game library. For .ini-style parsers, the only library I could find which was packaged for ubuntu, I posted on the daily wtf (it calculates an array of hash values, then does a linear scan through it...)

C++

Template syntax is heavy and unweildy. Let's see, for(map<string, int>::const_iterator it = mymap.begin(); it != mymap.end(); ++it). Design errors in the STL. Should changing allocation strategy for your vector really change its type? Overly complex type system. Type T1 has a convert-to-T2 method, and T2 has an implicit from-T1 constructor. Which is called? How does overloading, overriding and multiple inheritance interact? Poorly, I guess... Incredibly long and unwieldy error messages from templates. You know what I mean... References means you can't see output parameters at call sites. In C, you can guess what foo(bar, &baz) can and can't modify.

Python

No statements in lambdas. GRRRR foo( a for b in c if d ) feels wrong, it surprises me every time I get away with it. Shouldin't it be foo( (a for b in c if d) )? Can i have a dict comprehension? map and filter operators have special syntax in list comprehensions, how about something for reduce? or sort? Just by having a yield statement in it, a function is magically transformed into a generator, and its interface changes completely. Also, that generator cannot do any work before the first next(). at least, not without using a function that returns a generator.

JavaScript

No brief syntax for making modular code libraries. You have to call a function that returns a dictionary of public methods. And you have to edit that in (at least) two places every time you alter the interface of your module. Creating closures involves returning it from a function that returns a function from ('sup dog) yo' function. Clutter! for each ( foo ) syntax and behavior feels like an afterthought. Knowing when your code will actually run (and in what order) is more of a dark-art. The only way to get it right for sure is put everything (yes, that too) in one big file. and even then you still need to wait for a document.onload Am i missing something? is there no trivial way to get json serialized values without building them by hand? (yes jQuery can do this, sort of).

C#

Reference types are nullable by default; in-language null keyword is untyped. Lack of discriminated unions Exceptions as default, non-exceptional error handling method - there's not much of an alternative. archaic switch statement syntax and limitations Needless distinction between constructors + static methods Static methods can't be part of an interface Lack of by-shape interface implementation rather than explicit interface implementation - leading to numerous language design hacks such as the linq query syntax, foreach, collection & object initializers -- none of which can be flexibly reused. For example, the object initializer syntax may be nice, but plays poorly with immutable objects. Cannot inherit "interface" of a class independently of implementation - leading to code duplications and overarchitected code that provides interfaces, abstract base classes, a few common implementations, and no way to pick and choose the bits of each to use. Also; leads to too many code that's tightly coupled to a particular implementation since it's common to explicitly refer to the implementation type rather than an interface. Cannot multiply inherit via composition since a classes "interface" is tightly coupled to it's implementation; effectively lack of mixins. The above limitations of interfaces lead to a proliferation of virtually identical interfaces that don't overlap naturally in any kind of type hierarchy. IComparable vs. IEquatable vs. IComparable<T> vs object.Equals vs. operator == etc. etc. By extension, making a custom type that satisfies all these things is a lot more work than necessary (in particular for collection classes). Obviously, the language designers realize this, hence the various workarounds for things like linq, foreach and collection initializers which work by-shape rather than by-interface. Redundant use of parentheses and braces rather than layout-is-structure. Return values can be ignored, limiting the effectiveness of type inference. Enums aren't a normal type and can't have methods. Also, enum values aren't typesafe and may be initialized to 0 despite not having a 0 value. Mixing metaphors by lumping flag and non-flag enums together. Lack of proper value type support. Value types can't be inherited, have different constructor semantics, and perform poorly due to CLR limitations. Also, confusing semantics regarding value types: some values are really values (and can't be modified), and others are really non-aliased, non-null references (variables). This gets particularly confusing with regards to the next issue: Semantic distinction between fields and properties, particularly in conjunction with lack of mutability modifier (ala C++'s const) Can't specialize generics Cannot provide default generic type parameters (e.g. factory generics) lack of typedef makes generics a pain to use (using is a limited but good-to-know substitute!) Can't genericize over things other than types (e.g. functions, plain values, or names). This means you can't do something like make a generic implementation of a dependancy property leading to, well, nasty implementations of things like dependancy properties and the overuse of code-snippets and poorly readable code as a result. Limited capability to specify generic type requirements e.g. generic sum method that takes both int, double and a bigint (without tricky and often slow hacks). An interface method implementation or virtual method override cannot return a more specific type or accept a more general type; i.e. limited co/contravariance support even in C# 4.

Haskell(包括所有GHC扩展，而不仅仅是Haskell'98的基础规范)。

我讨厌它的一点是:它不是主流。