What is the difference between yield and yield from?
yield from delegates the full iterator protocol to a sub-iterator
yield from iterable is generator delegation introduced by PEP 380. It does not just loop: it forwards send(), throw(), and close() to the sub-iterator, and the value the sub-iterator returns via StopIteration becomes the value of the yield from expression. A manual for loop over the sub-iterator yields the same items but silently drops all of that: send() values never reach the inner generator, exceptions are not thrown into it, and its return value is lost. So yield from is the correct primitive when you are delegating to a sub-generator, and a for loop is correct when you want to transform each item.
- 1
Use yield from for flattening, recursive generator walkers, and any delegation where the inner generator may be sent values or has a meaningful return value.
- 2
Use an explicit for loop with yield when you need to inspect, filter, or transform each item, or when you want to instrument the flow.
- 3
Trade-off: yield from makes the delegation opaque, which is great for correctness and bad for per-item logging or profiling.
- 4
Common mistake: writing a for loop wrapper around a coroutine-style generator and expecting send() to work through it. It will not.
- 5
Common mistake: assuming yield from accepts any iterable for send purposes; delegation only forwards send/throw when the sub-iterator supports them.
- 6
Version note: yield from arrived in Python 3.3 (PEP 380). It is a SyntaxError inside async def functions and async generators, where await and async for are the equivalents.
Scenario Questions
0-2 years experience
- 1You need to flatten a list of lists lazily. How does yield from simplify the code?
- 2What value does the expression 'yield from xs' evaluate to?
2-5 years experience
- 1You wrap a sub-generator in a for loop instead of yield from. What behavior do you lose?
- 2How do you retrieve the return value of a sub-generator that finishes with return?
5-8 years experience
- 1You are building a recursive tree walker with generators. Why is yield from usually preferred over manual recursion?
- 2You need per-item instrumentation across a nested generator chain. Why might yield from make that harder, and how do you work around it?
8+ years experience
- 1Design a delegation layer that correctly forwards close() and throw() into a pool of sub-generators and handles partial consumption.
- 2You are porting generator-based coroutine code to async/await. What does yield from map to and where does the mapping break down?
Follow-up Questions
- How does a manual for-loop delegation break a generator that expects send()?
- Why is yield from a SyntaxError inside an async def, and what replaces it?