pipe into stdin vs attach -- differences?

edit

Here's some code, called slopes.sage:

def classical_slopes(N,p,r,k,i):
    assert p>2
    assert r>1 ## wild
    G=DirichletGroup(p^r) ## cyclic
    alpha=G[1]
    alpha_tame=alpha^(p^(r-1))
    alpha_wild=alpha^(p-1)
    K=alpha.base_ring() ## cyclotomic
    chi=alpha_tame^i*alpha_wild
    C=CuspForms(chi.extend(N*p^r),k)
    f=C.hecke_polynomial(p).base_extend(K)
    R=K.maximal_order()
    primroot=primitive_root(p) ## p>2
    X=R.ideal(p,alpha_tame(primroot)-primroot).factor()
    assert len(X)==1 ## ideal should be a prime power
    Qt = PolynomialRing(QQ, 't')
    P,e=X[0]
    c2=[2^(a.valuation(P)) if a!=0 else 0 for a in f.coefficients()]
    return (e,Qt(c2).newton_slopes(2))

# off we go. The import sys bit is to flush stdout.
import sys
N=3
p=5
for k in range(2,10):
    i=4-(k%4)
    print N,p,3,k,i,classical_slopes(N,p,3,k,i)
    sys.stdout.flush()

If I pipe it into sage like this:

sage < slopes.sage

I get the following output:

$ sage < slopes.sage 
┌────────────────────────────────────────────────────────────────────┐
│ Sage Version 6.5, Release Date: 2015-02-17                         │
│ Type "notebook()" for the browser-based notebook interface.        │
│ Type "help()" for help.                                            │
└────────────────────────────────────────────────────────────────────┘

sage: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

....: 

sage: 

sage: 

sage: 

sage: 

sage: 

....: 

....: 

....: 

....: 

Exiting Sage (CPU time 0m0.03s, Wall time 0m0.20s).

But if I attach the file, within a sage console session, it runs as I expect it to run. Is this something to do with the pre-parser? If I've made a slip then that's great, but if it's harder than I think to pipe sage script into sage then I'd appreciate some tips as I would like to start a large sage job on a remote machine via ssh.