Use of RecursionError breaks compatibility with Python 2.7 and 3.4

elasticsearch-py version: 7.13.2

Description of the problem including expected versus actual behavior:

In commit cfdfd5111fd, use of the RecursionError class was introduced. This class was added in Python 3.5, though.

This causes errors as this under Python 2.7:

    def perform_request(
        self, method, url, params=None, body=None, timeout=None, ignore=(), headers=None
    ):
        url = self.url_prefix + url
        if params:
            url = "%s?%s" % (url, urlencode(params))
    
        full_url = self.host + url
    
        start = time.time()
        orig_body = body
        try:
            kw = {}
            if timeout:
                kw["timeout"] = timeout
    
            # in python2 we need to make sure the url and method are not
            # unicode. Otherwise the body will be decoded into unicode too and
            # that will fail (#133, #201).
            if not isinstance(url, str):
                url = url.encode("utf-8")
            if not isinstance(method, str):
                method = method.encode("utf-8")
    
            request_headers = self.headers.copy()
            request_headers.update(headers or ())
    
            if self.http_compress and body:
                body = self._gzip_compress(body)
                request_headers["content-encoding"] = "gzip"
    
            response = self.pool.urlopen(
                method, url, body, retries=Retry(False), headers=request_headers, **kw
            )
            duration = time.time() - start
            raw_data = response.data.decode("utf-8", "surrogatepass")
>       except RecursionError:
E       NameError: global name 'RecursionError' is not defined

../../../.venv/py2/lib/python2.7/site-packages/elasticsearch/connection/http_urllib3.py:256: NameError

According to #1295 and setup.py, Python 2.7 and 3.4 should still be supported, though.

[MarcSchmitzer]

Thanks for looking into this. I guess this hasn’t shown up in the tests because the specific error is not triggered. As far as I understand, the except expression is not evaluated unless there actually is an exception to handle.

The error is actually easy to trigger by trying to talk to a non-existent server:

from elasticsearch.client import Elasticsearch
client = Elasticsearch("http://localhost:9", max_retries=0)  # nothing listens on this port
client.indices.stats("_all")

[sethmlarson]

Closed in https://github.com/elastic/elasticsearch-py/pull/1609