Infrastructure

Reusing a connection with DataONEBaseClient will fail if the subsequent request is not sent within a few seconds since the server may have closed the connection. This is manifest in an exception "httplib.BadStatusLine", which is really just a symptom of the actual problem, which is that the socket is no longer in a state for writing.

The solution is one of:

1. call connection.close() after each request

2. trap the exception, close the connection, and reissue the call

3. check the socket state before issuing the request

Of these #2 seems most reasonable, since 1. #1 is wasteful, and for 3. #3 the state may have changed between the call to socket.select and the request is made.

To reproduce the issue:

<pre>

import select
import time
import httplib
import logging
import traceback
from d1_client import d1baseclient

logging.basicConfig(level=logging.DEBUG)
burl = "https://cn.dataone.org/cn"
c = d1baseclient.DataONEBaseClient(burl)
t0 = time.time()
ol = c.listObjects(start=0, count=100)
fileno = c.connection.sock.fileno()
sockok = True
while sockok:
res = select.select([fileno],[fileno],[fileno],0)
logging.debug("select response: %s" % str(res) )
sockok = res[0] == []
time.sleep(0.5)
logging.info( "Delta = %.2f" % (time.time() - t0) )
try:
ol = c.listObjects(start=0, count=100)
except httplib.BadStatusLine as e:
logging.error(traceback.format_exc())
c.connection.close()
ol = c.listObjects(start=0, count=100)

</pre>