Hamming numbers

Haskell, 101 97 92+|n| characters

h=1:m 2h&m 3h&m 5h
m=map.(*)
c@(a:b)&o@(m:n)|a<m=a:b&o|a>m=m:c&n|0<1=a:b&n
main=print$take 1000000h

Computes the full million in 3.7s on the machine I tested on (variably more if you actually want the output stored)

Ungolfed:

-- print out the first million Hamming numbers
main = print $ take 1000000 h

-- h is the entire Hamming sequence.
-- It starts with 1; for each number in the
-- sequence, 2n, 3n and 5n are also in.
h = 1 : (m 2 h) & (m 3 h) & (m 5 h)

-- helper: m scales a list by a constant factor
m f xs = map (f*) xs

-- helper: (&) merges two ordered sequences
a@(ha:ta) & b@(hb:tb)
    |    ha < hb = ha : ta & b
    |    ha > hb = hb :  a & tb
    |  otherwise = ha : ta & tb

All Haskell is notoriously good at: defining a list as a lazy function of itself, in a way that actually works.

Python 181 Characters

h=[]        
h.append(1)
n=input()
i=j=k=0
while n:
    print h[-1]
    while h[i]*2<=h[-1]:
        i+=1
    while h[j]*3<=h[-1]:
        j+=1
    while h[k]*5<=h[-1]:
        k+=1
    h.append(min(h[i]*2,h[j]*3,h[k]*5))
    n-=1

APL (Dyalog Classic), 34 23 bytes

{⍺⍴{⍵[⍋⍵]}∪,⍵∘.×⍳5}⍣≡∘1

Try it online!

TIO throws a WS FULL error for \$n = 1000000\$, but Dyalog on my laptop runs in about 45 seconds, not counting the scrolling to display numbers.

{⍺⍴{⍵[⍋⍵]}∪,⍵∘.×⍳5}⍣≡∘1     Monadic function:
{⍺⍴{⍵[⍋⍵]}∪,⍵∘.×⍳5}         Define the following helper function g(⍺,⍵):
             ⍵∘.×⍳5             Make a multiplication table between ⍵ and (1 2 3 4 5).
                                (Including 4 is unnecessary but saves bytes.)
            ,                   Flatten the table into an array.
           ∪                    Keep unique elements.
    {⍵[⍋⍵]}                     Grade up the array and access it at those indices.
                                (This is the APL idiom to sort an array.)
 ⍺⍴                             Keep the first ⍺ elements; pad by repeating the array.
{⍺⍴{⍵[⍋⍵]}∪,⍵∘.×⍳5}⍣≡       Repeatedly apply g with some fixed left argument
                             until a fixed point is reached.
                             At this point we have a dyadic function that takes
                             n on the left and the starting value on the right,
                             and returns multiples of the n Hamming numbers.
                      ∘1     Fix 1 as the right argument.