Checking for EOF in stream

[GiBo]

Hi!

Classic situation - I have to process an input stream of unknown length
until a I reach its end (EOF, End Of File). How do I check for EOF? The
input stream can be anything from opened file through sys.stdin to a
network socket. And it's binary and potentially huge (gigabytes), thus
"for line in stream.readlines()" isn't really a way to go.

For now I have roughly:

stream = sys.stdin
while True:
data = stream.read(1024)
process_data(data)
if len(data) < 1024: ## (*)
break

I smell a fragile point at (*) because as far as I know e.g. network
sockets streams may return less data than requested even when the socket
is still open.

I'd better like something like:

while not stream.eof():
...

but there is not eof() method :-(

This is probably a trivial problem but I haven't found a decent solution.

Any hints?

Thanks!

GiBo

 

[Grant Edwards]

Hi!

Classic situation - I have to process an input stream of unknown length
until a I reach its end (EOF, End Of File). How do I check for EOF? The
input stream can be anything from opened file through sys.stdin to a
network socket. And it's binary and potentially huge (gigabytes), thus
"for line in stream.readlines()" isn't really a way to go.

For now I have roughly:

stream = sys.stdin
while True:
data = stream.read(1024)

if len(data) == 0:
break #EOF

 

[GiBo]

if len(data) == 0:
break #EOF

Right, not a big difference though. Isn't there a cleaner / more
intuitive way? Like using some wrapper objects around the streams or
something?

GiBo

 

[Gabriel Genellina]

Right, not a big difference though. Isn't there a cleaner / more
intuitive way? Like using some wrapper objects around the streams or
something?

Read the documentation... For a true file object:
read([size]) ... An empty string is returned when EOF is encountered
immediately.
All the other "file-like" objects (like StringIO, socket.makefile, etc)
maintain this behavior.
So this is the way to check for EOF. If you don't like how it was spelled,
try this:

if data=="": break

If your data is made of lines of text, you can use the file as its own
iterator, yielding lines:

for line in stream:
process_line(line)

 

[Grant Edwards]

Right, not a big difference though. Isn't there a cleaner /
more intuitive way?

A file is at EOF when read() returns ''. The above is the
cleanest, simplest, most direct way to do what you specified.
Everybody does it that way, and everybody recognizes what's
being done.

It's also the "standard, Pythonic" way to do it.

Like using some wrapper objects around the streams or
something?

You can do that, but then you're mostly just obfuscating
things.

 

[Jon Ribbens]

So this is the way to check for EOF. If you don't like how it was spelled,
try this:

if data=="": break

How about:

if not data: break

? ;-)

 

[Nathan]

Right, not a big difference though. Isn't there a cleaner / more
intuitive way? Like using some wrapper objects around the streams or
something?

Read the documentation... For a true file object:
read([size]) ... An empty string is returned when EOF is encountered
immediately.
All the other "file-like" objects (like StringIO, socket.makefile, etc)
maintain this behavior.
So this is the way to check for EOF. If you don't like how it was spelled,
try this:

if data=="": break

If your data is made of lines of text, you can use the file as its own
iterator, yielding lines:

for line in stream:
process_line(line)

Not to beat a dead horse, but I often do this:

data = f.read(bufsize):
while data:
# ... process data.
data = f.read(bufsize)


-The only annoying bit it the duplicated line. I find I often follow
this pattern, and I realize python doesn't plan to have any sort of
do-while construct, but even still I prefer this idiom. What's the
concensus here?

What about creating a standard binary-file iterator:

def blocks_of(infile, bufsize = 1024):
data = infile.read(bufsize)
if data:
yield data


-the use would look like this:

for block in blocks_of(myfile, bufsize = 2**16):
process_data(block) # len(block) <= bufsize...

 

[Nathan]

Grant Edwards wrote:

Classic situation - I have to process an input stream of unknown length
until a I reach its end (EOF, End Of File). How do I check for EOF? The
input stream can be anything from opened file through sys.stdin to a
network socket. And it's binary and potentially huge (gigabytes), thus
"for line in stream.readlines()" isn't really a way to go.

For now I have roughly:

stream = sys.stdin
while True:
data = stream.read(1024)
if len(data) == 0:
break #EOF
process_data(data)

Right, not a big difference though. Isn't there a cleaner / more
intuitive way? Like using some wrapper objects around the streams or
something?

Read the documentation... For a true file object:
read([size]) ... An empty string is returned when EOF is encountered
immediately.
All the other "file-like" objects (like StringIO, socket.makefile, etc)
maintain this behavior.
So this is the way to check for EOF. If you don't like how it was spelled,
try this:

if data=="": break

If your data is made of lines of text, you can use the file as its own
iterator, yielding lines:

for line in stream:
process_line(line)

Not to beat a dead horse, but I often do this:

data = f.read(bufsize):
while data:
# ... process data.
data = f.read(bufsize)


-The only annoying bit it the duplicated line. I find I often follow
this pattern, and I realize python doesn't plan to have any sort of
do-while construct, but even still I prefer this idiom. What's the
concensus here?

What about creating a standard binary-file iterator:

def blocks_of(infile, bufsize = 1024):
data = infile.read(bufsize)
if data:
yield data


-the use would look like this:

for block in blocks_of(myfile, bufsize = 2**16):
process_data(block) # len(block) <= bufsize...

(ahem), make that iterator something that works, like:

def blocks_of(infile, bufsize = 1024):
data = infile.read(bufsize)
while data:
yield data
data = infile.read(bufsize)

 

[kousue]

Hi!

Classic situation - I have to process an input stream of unknown length
until a I reach its end (EOF, End Of File). How do I check for EOF? The
input stream can be anything from opened file through sys.stdin to a
network socket. And it's binary and potentially huge (gigabytes), thus
"for line in stream.readlines()" isn't really a way to go.

Could you use xreadlines()? It's a lazily-evaluated stream reader.

For now I have roughly:

stream = sys.stdin
while True:
data = stream.read(1024)
process_data(data)
if len(data) < 1024: ## (*)
break

I smell a fragile point at (*) because as far as I know e.g. network
sockets streams may return less data than requested even when the socket
is still open.

Well it depends on a lot of things. Is the stream blocking or non-
blocking (on sockets and some other sorts of streams, you can pick
this yourself)? What are the underlying semantics (reliable-and-
blocking TCP or dropping-and-unordered-UDP)? Unfortunately, you really
need to just know what you're working with (and there's really no
better solution; trying to hide the underlying semantics under a
proscribed overlaid set of semantics can only lead to badness in the
long run).

I'd better like something like:

while not stream.eof():
...

but there is not eof() method :-(

This is probably a trivial problem but I haven't found a decent solution.

For your case, it's not so hard:
http://pyref.infogami.com/EOFError says "read() and readline() methods
of file objects return an empty string when they hit EOF." so you
should assume that if something is claiming to be a file-like object
that it will work this way.

Any hints?

So:
stream = sys.stdin
while True:
data = stream.read(1024)
if data=="":
break
process_data(data)