MemoryError

Python MemoryError on Windows often traces to an overlooked installation detail: 32-bit Python has a hard addressable memory ceiling of approximately 2 GB per process, regardless of how much RAM the machine has. Developers who downloaded the 32-bit Python installer instead of the 64-bit one, and who process large datasets, hit MemoryError on machines with 16 GB RAM — because the 32-bit Python process cannot address memory above the 2 GB virtual address space limit. Run import platform; print(platform.architecture()) to confirm. The output ('32bit', 'WindowsPE') confirms you need to reinstall the 64-bit version. NumPy operations are the most common trigger: allocating a numpy.zeros((10000, 10000), dtype=float64) array requires approximately 800 MB — straightforward on 64-bit Python, impossible on 32-bit. For 64-bit Python processes genuinely running out of memory on large datasets, tracemalloc is the built-in profiling tool: tracemalloc.start() before your code, then tracemalloc.take_snapshot().statistics('lineno') after identifies which lines allocate the most memory. Processing data in chunks using generators (yield) instead of loading entire datasets into memory is the standard mitigation. The pandas chunksize parameter for read_csv() processes CSV files in batches without loading the full file, reducing peak memory from the file size to the chunk size.