TODO

-*-outline-*-

* doc, docstrings

* more tests! (See coverage)

* use tags to distinguish tests that require net connectivity from others
Net tests need to not use cache (easier when these are separated! -- #123)
But... commit cache into repo for non-net tests

* distinguish tests that use vcr-cached data (common) from live tests that
  ensure that the package still works against current endpoints, which obviously
  need to be less stringent about exact matches.

* filter out unused headers in cassettes

* reenable testing for dbsnp xmlfacade; see tests/x-test_eutils_xmlfacades_dbsnp.py

* switch to pytest-vcr config

* FIX xmlfacade for dbsnp

* #122: implement TTL support in cache

* #123: refactor queryservice to separate querying, throttling, and caching
queryservice implements too much functionality. It should be refactored into:
a basic query service that executes queries, period
a thin throttling service based on the basic query service
a caching service on the throttled service (since that's all the caching service should use)
separate caching, throttling, querying, parsing/facades

* #124: Support large search result sets
support webenv history
use web history + iterators for chained searching

* add test tags to isolate tests against live ncbi data (ie not pyvcr tests)

* QueryService:
** Fetch & compare
** TTL support in cache, request-specific TTLs?
** optional db -> options map (esp. for rettype & retmode)
** deal with caching status 200 replies that are bogus (e.g., truncated xml) -- callbacks?
** provide uncached access
** support history
** default args is misplaced -- it should go in client instead