2014-10-05 21:44:31 +00:00
|
|
|
.. _example_resolve_name:
|
|
|
|
|
|
|
|
Resolve a name
|
2017-06-16 10:16:05 +00:00
|
|
|
==============
|
2014-10-05 21:44:31 +00:00
|
|
|
|
2017-06-16 10:16:05 +00:00
|
|
|
This basic example shows how to create a context and resolve a host address
|
|
|
|
(DNS record of A type).
|
|
|
|
|
|
|
|
Source code
|
|
|
|
-----------
|
2014-10-05 21:44:31 +00:00
|
|
|
|
|
|
|
::
|
|
|
|
|
2017-06-16 10:16:05 +00:00
|
|
|
#!/usr/bin/python
|
|
|
|
import unbound
|
|
|
|
|
|
|
|
ctx = unbound.ub_ctx()
|
|
|
|
ctx.resolvconf("/etc/resolv.conf")
|
|
|
|
|
|
|
|
status, result = ctx.resolve("www.google.com")
|
|
|
|
if status == 0 and result.havedata:
|
|
|
|
print "Result.data:", result.data.address_list
|
|
|
|
elif status != 0:
|
|
|
|
print "Resolve error:", unbound.ub_strerror(status)
|
|
|
|
|
|
|
|
In contrast with the C API, the source code is more compact while the
|
|
|
|
performance of C implementation is preserved.
|
|
|
|
The main advantage is that you need not take care about the deallocation and
|
|
|
|
allocation of context and result structures; pyUnbound module does it
|
|
|
|
automatically for you.
|
|
|
|
|
|
|
|
If only domain name is given, the :meth:`unbound.ub_ctx.resolve` looks for
|
|
|
|
A records in IN class.
|