Generator Expressions
As I learned more about Python’s iterator protocol and the different ways to implement it in my own code, I realized that “syntactic sugar” was a recurring theme.
You see, class-based iterators and generator functions are two expressions of the same underlying design pattern.
Generator functions give you a shortcut for supporting the iterator protocol in your own code, and they avoid much of the verbosity of class-based iterators. With a little bit of specialized syntax, or syntactic sugar, they save you time and make your life as a developer easier.
This is a recurring theme in Python and in other programming languages. As more developers use a design pattern in their programs, there’s a growing incentive for the language creators to provide abstractions and implementation shortcuts for it.
That’s how programming languages evolve over time—and as developers, we reap the benefits. We get to work with more and more powerful building blocks, which reduces busywork and lets us achieve more in less time.
Earlier in this book you saw how generators provide syntactic sugar for writing class-based iterators. The generator expressions we’ll cover in this chapter add another layer of syntactic sugar on top.
Generator expressions give you an even more effective shortcut for writing iterators. With a simple and concise syntax that looks like a list comprehension, you’ll be able to define iterators in a single line of code.
Here’s an example:
iterator = ('Hello' for i in range(3))
When iterated over, this generator expression yields the same sequence of values as the bounded_repeater generator function we wrote in the previous chapter. Here it is again to refresh your memory:
def bounded_repeater(value, max_repeats):
for i in range(max_repeats):
yield value
iterator = bounded_repeater('Hello', 3)
Isn’t it amazing how a single-line generator expression now does a job that previously required a four-line generator function or a much longer class-based iterator?
But I’m getting ahead of myself. Let’s make sure our iterator defined with a generator expression actually works as expected:
>>> iterator = ('Hello' for i in range(3))
>>> for x in iterator:
... print(x)
'Hello'
'Hello'
'Hello'
That looks pretty good to me! We seem to get the same results from our one-line generator expression that we got from the bounded_repeater generator function.
There’s one small caveat though: Once a generator expression has been consumed, it can’t be restarted or reused. So in some cases there is an advantage to using generator functions or class-based iterators.
Generator Expressions vs List Comprehensions
As you can tell, generator expressions are somewhat similar to list comprehensions:
>>> listcomp = ['Hello' for i in range(3)]
>>> genexpr = ('Hello' for i in range(3))
Unlike list comprehensions, however, generator expressions don’t construct list objects. Instead, they generate values “just in time” like a class-based iterator or generator function would.
All you get by assigning a generator expression to a variable is an iterable “generator object”:
>>> listcomp
['Hello', 'Hello', 'Hello']
>>> genexpr
<generator object <genexpr> at 0x1036c3200>
To access the values produced by the generator expression, you need to call next() on it, just like you would with any other iterator:
>>> next(genexpr)
'Hello'
>>> next(genexpr)
'Hello'
>>> next(genexpr)
'Hello'
>>> next(genexpr)
StopIteration
Alternatively, you can also call the list() function on a generator expression to construct a list object holding all generated values:
>>> genexpr = ('Hello' for i in range(3))
>>> list(genexpr)
['Hello', 'Hello', 'Hello']
Of course, this was just a toy example to show how you can “convert” a generator expression (or any other iterator for that matter) into a list. If you need a list object right away, you’d normally just write a list comprehension from the get-go.
Let’s take a closer look at the syntactic structure of this simple generator expression. The pattern you should begin to see looks like this:
genexpr = (expression for item in collection)
The above generator expression “template” corresponds to the following
generator function:
def generator():
for item in collection:
yield expression
Just like with list comprehensions, this gives you a “cookie-cutter pattern” you can apply to many generator functions in order to transform them into concise generator expressions.
Filtering Values
There’s one more useful addition we can make to this template, and that’s element filtering with conditions. Here’s an example:
>>> even_squares = (x * x for x in range(10) if x % 2 == 0)
This generator yields the square numbers of all even integers from zero to nine. The filtering condition using the % (modulo) operator will reject any value not divisible by two:
>>> for x in even_squares:
... print(x)
0
4
16
36
64
Let’s update our generator expression template. After adding element filtering via if-conditions, the template now looks like this:
genexpr = (expression for item in collection if condition)
And once again, this pattern corresponds to a relatively straightforward, but longer, generator function. Syntactic sugar at its best:
def generator():
for item in collection:
if condition:
yield expression
In-line Generator Expressions
Because generator expressions are, well…expressions, you can use them in-line with other statements. For example, you can define an iterator and consume it right away with a for-loop:
for x in ('Bom dia' for i in range(3)):
print(x)
There’s another syntactic trick you can use to make your generator expressions more beautiful. The parentheses surrounding a generator expression can be dropped if the generator expression is used as the single argument to a function:
>>> sum((x * 2 for x in range(10)))
90
# Versus:
>>> sum(x * 2 for x in range(10))
90
This allows you to write concise and performant code. Because generator expressions generate values “just in time” like a class-based iterator or a generator function would, they are very memory efficient.
Too Much of a Good Thing…
Like list comprehensions, generator expressions allow for more complexity than what we’ve covered so far. Through nested for-loops and chained filtering clauses, they can cover a wider range of use cases:
(expr for x in xs if cond1
for y in ys if cond2
...
for z in zs if condN)
The above pattern translates to the following generator function logic:
for x in xs:
if cond1:
for y in ys:
if cond2:
...
for z in zs:
if condN:
yield expr
And this is where I’d like to place a big caveat:
Please don’t write deeply nested generator expressions like that. They can be very difficult to maintain in the long run.
难以维护。
This is one of those “the dose makes the poison” situations where a beautiful and simple tool can be overused to create hard to read and difficult to debug programs.
Just like with list comprehensions, I personally try to stay away from any generator expression that includes more than two levels of nesting.
两级内嵌后就要注意了。
Generator expressions are a helpful and Pythonic tool in your toolbox, but that doesn’t mean they should be used for every single problem you’re facing. For complex iterators, it’s often better to write a generator function or even a class-based iterator.
If you need to use nested generators and complex filtering conditions, it’s usually better to factor out sub-generators (so you can name them) and then to chain them together again at the top level. You’ll see how to do this in the next chapter on iterator chains.
If you’re on the fence, try out different implementations and then select the one that seems the most readable. Trust me, it’ll save you time in the long run.
Key Takeaways
- Generator expressions are similar to list comprehensions. However, they don’t construct list objects. Instead, generator expressions generate values “just in time” like a class-based iterator or generator function would.
- Once a generator expression has been consumed, it can’t be restarted or reused.
- Generator expressions are best for implementing simple “ad hoc” iterators. For complex iterators, it’s better to write a generator function or a class-based iterator.
