# Find Median from Data Stream Two Heaps (Practice 5078): Sketch
Interviewers use this pattern to test whether you can state the invariant before coding. Here is a compact rehearsal note.
## Approach
A max-heap for the low half and a min-heap for the high half, rebalanced after every insert.
```python
import heapq
class MedianFinder:
def __init__(self): self.lo=[];self.hi=[]
def add(self,x):
heapq.heappush(self.lo,-x);heapq.heappush(self.hi,-heapq.heappop(self.lo))
if len(self.hi)>len(self.lo): heapq.heappush(self.lo,-heapq.heappop(self.hi))
def median(self):
return -self.lo[0] if len(self.lo)>len(self.hi) else (-self.lo[0]+self.hi[0])/2
```
## Complexity
Linear in the dominant input size for the scan; heap/tree variants add a log factor.
## Interview pitfalls
State the invariant out loud, name the complexity, then code. Skipping the invariant is the usual fail mode.
## Practicing out loud
Narrate the approach, complexity and one edge case in under a minute. Disclosure: I work on Linkjob AI, a real-time interview assistant. If you want live practice help, see Linkjob AI copilot.