printf format f versus d in Python

Question

Some C code I am working in has an int set to -145. Depending on the format, the value returns as either 0.00 or -145. Take the following

int a = -145; printf("%.2lf", a); // returns 0.00 printf("%.2ld", a); // returns -145

First question is how is it that the first line is returning 0.00? Second question is how to return 0.00 in Python?

Among the many iterations I've tried, here are a few:

a = -145 "%0.2lf" % a # -145.00 "%0.2ld" % a # -145 "%0.2li" % a # -145 "%0.2f" % a # -145.00

Of course @ouah is correct: C defines nothing about what happens in the code you showed; it will vary across platforms, and may result in anything from displaying gibberish to crashing with a segfault. C's printf family is notoriously prone to errors this way. To answer your 2nd question, you cannot "trick" Python this way - there's no way to get Python to treat -145 as anything other than what it is: an integer. — Tim Peters
– Tim Peters, Commented Dec 17, 2013 at 22:29

ouah · Accepted Answer · 2013-12-17 22:15:19Z

 int a = -145; printf("%.2lf", a); // returns 0.00 printf("%.2ld", a); // returns -145

Both function calls invoke undefined behavior. The correct conversion specifier to print an int is d not lf (for double), not ld (for long).

dan04 · Accepted Answer · 2013-12-18 00:13:36Z

Second question is how to return 0.00 in Python?

If you insist:

import struct def misinterpret_int_as_double(n): int_bytes = struct.pack('i', n) padding = b'\x00' * (struct.calcsize('d') - struct.calcsize('i')) return struct.unpack('d', int_bytes + padding)[0] >>> misinterpret_int_as_double(-145) 2.1219957193e-314

This is essentially what your C code is doing behind the scenes. C doesn't do any type-checking (at compile-time or run-time) on the arguments passed in the ... part of a call to a varargs function. What happens is that the memory that stores printf's arguments contains:

A pointer to the string literal "%.2lf.
The bytes representing the number -145. (On x86-32 or x86-64, this is the 4 bytes 91 FF FF FF.)
Some garbage data. (In the Python code above, this is assumed to be all zeros, but in your C program it need not be.)

The printf function sees the lf specifier and expects a double. So, it interprets the following bytes, 91 FF FF FF xx xx xx xx (where xx = garbage) as one. For about 1/4 of the possible values of the garbage (including 00 00 00 00), the number is small enough to round to zero.

Note that I've assumed a bunch of stuff: That you have a little-endian system with 4-byte int and 8-byte double. That function arguments are passed in ascending order in memory. And that your code doesn't segfault. YMMV on other hardware/OS/compiler combinations. That's how it is with undefined behavior.

Python works differently, because it's safely typed. If you pass the “wrong” type to str's % operator, it will automatically convert the operand to the correct type by calling the magic __float__ method (or __str__ or __int__ or whatever depending on which format it is).

Collectives™ on Stack Overflow

printf format f versus d in Python

2 Answers 2

Comments

Comments

Hot Network Questions

Collectives™ on Stack Overflow

2 Answers 2

Comments

Comments

Related