示例#1
0
文件: h2o.py 项目: h2oai/h2o-3
    def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
                          cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, suppressErrorMsg=False, **kwargs):
        H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
        # if url param is used, use it as full url. otherwise crate from the jsonRequest
        if fullUrl:
            url = fullUrl
        else:
            url = self.__url(jsonRequest)

        # remove any params that are 'None'
        # need to copy dictionary, since can't delete while iterating
        if params is not None:
            params_serialized = params.copy()
            for k in params_serialized:
                if params_serialized[k] is None:
                    del params[k]
            paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
        else:
            paramsStr = ''

        # The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
        # NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
        # NOTE: we currently don't need to do this for GET, so that's not implemented.
        if postData is not None:
            munged_postData = {}
            for k, v in postData.iteritems():
                if type(v) is list:
                    if len(v) == 0:
                        munged_postData[k] = '[]'
                    else:
                        first = True
                        array_str = '['
                        for val in v:
                            if not first: array_str += ', '

                            if val is None:
                                array_str += 'null'
                            elif isinstance(val, basestring):
                                array_str += "\"" + str(val) + "\""
                            else:
                                array_str += str(val)
                            first  = False
                        array_str += ']'
                        munged_postData[k] = array_str
                elif type(v) is dict:
                    if len(v) == 0:
                        munged_postData[k] = '{}'
                    else:
                        first = True
                        map_str = '{'
                        for key, val in v.iteritems():
                            if not first: map_str += ', '

                            if val is None:
                                map_str += "\"" + key + "\"" + ': null'
                            elif isinstance(val, basestring):
                                map_str += "\"" + str(key) + "\"" + ":" + "\"" + str(val) + "\""
                            else:
                                map_str += "\"" + key + "\"" + ':' + str(val)
                            first  = False
                        map_str += '}'
                        munged_postData[k] = map_str

                else:
                    # not list:
                    munged_postData[k] = v
        else:
            # None
            munged_postData = postData

        # print("munged_postData: " + repr(munged_postData))

        if extraComment:
            log('Start ' + url + paramsStr, comment=extraComment)
        else:
            log('Start ' + url + paramsStr)

        log_rest("")
        log_rest("----------------------------------------------------------------------\n")
        if extraComment:
            log_rest("# Extra comment info about this request: " + extraComment)
        if cmd == 'get':
            log_rest("GET")
        else:
            log_rest("POST")
        log_rest(url + paramsStr)

        # file get passed thru kwargs here
        try:
            if 'post' == cmd:
                # NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
                # This is temporary.
                #
                # This following does application/json (aka, posting JSON in the body):
                # r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
                #
                # This does form-encoded, which doesn't allow POST of nested structures
                r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
            elif 'delete' == cmd:
                r = requests.delete(url, timeout=timeout, params=params, **kwargs)
            elif 'get' == cmd:
                r = requests.get(url, timeout=timeout, params=params, **kwargs)
            else:
                raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)

        except Exception as e:
            # rethrow the exception after we've checked for stack trace from h2o
            # out of memory errors maybe don't show up right away? so we should wait for h2o
            # to get it out to h2o stdout. We don't want to rely on cloud teardown to check
            # because there's no delay, and we don't want to delay all cloud teardowns by waiting.
            # (this is new/experimental)
            exc_info = sys.exc_info()
            # use this to ignore the initial connection errors during build cloud when h2o is coming up
            if not noExtraErrorCheck:
                h2p.red_print(
                    "ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
                time.sleep(2)
                H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
            log_rest("")
            log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
            raise (exc_info[1], None, exc_info[2])

            H2O.verboseprint("r: " + repr(r))

        if 200 != r.status_code:
            pp = pprint.PrettyPrinter(indent=4)
            msg = "JSON call returned non-200 status: " + url

            json = r.json()
            if None != json and 'dev_msg' in json:
                msg += "\ndev_msg: "
                msg += str(json['dev_msg'])
            msg += "\nr.status_code: " + str(r.status_code)
            msg += "\nr.headers: " + repr(r.headers)
            if None == json:
                msg += '\nERROR: the error output from H2O is not JSON!'
                msg += "\nr.text: " + r.text
            else:
                msg += "\nr.json: "
                msg += pp.pformat(json)

            if raiseIfNon200:
                pass  # we'll pass msg up with the exception
            elif not suppressErrorMsg:
                print(msg)
            log_rest(msg)

        log_rest("")
        try:
            if r is None:
                log_rest("r is None")
            else:
                log_rest("HTTP status code: " + str(r.status_code))
                # The following accesses to r.text were taking most of the runtime:
                log_text = False
                if log_text:
                    if hasattr(r, 'text'):
                        if r.text is None:
                            log_rest("r.text is None")
                        else:
                            log_rest(r.text)
                    else:
                        log_rest("r does not have attr text")
        except Exception as e:
            # Paranoid exception catch.
            # Ignore logging exceptions in the case that the above error checking isn't sufficient.
            print("Caught exception from result logging: ", e, "; result: ", repr(r))

        # fatal if no response
        if raiseIfNon200 and not r:
            raise Exception("Maybe bad url? no r in __do_json_request in %s:" % inspect.stack()[1][3] + "\n\n" + msg)

        # this is used to open a browser on results, or to redo the operation in the browser
        # we don't' have that may urls flying around, so let's keep them all
        H2O.json_url_history.append(r.url)
        # if r.json():
        #     raise Exception("Maybe bad url? no r.json in __do_json_request in %s:" % inspect.stack()[1][3])

        rjson = None
        if returnFast:
            return
        try:
            rjson = r.json()
        except:
            print(h2o_test_utils.dump_json(r.text))
            if not isinstance(r, (list, dict)):
                raise Exception("h2o json responses should always be lists or dicts, see previous for text")

            raise Exception("Could not decode any json from the request.")

        # TODO
        # TODO
        # TODO
        # TODO: we should really only look in the response object.  This check
        # prevents us from having a field called "error" (e.g., for a scoring result).
        for e in ['error', 'Error', 'errors', 'Errors']:
            # error can be null (python None). This happens in exec2
            if e in rjson and rjson[e]:
                H2O.verboseprint("rjson:" + h2o_test_utils.dump_json(rjson))
                emsg = 'rjson %s in %s: %s' % (e, inspect.stack()[1][3], rjson[e])
                if ignoreH2oError:
                    # well, we print it..so not totally ignore. test can look at rjson returned
                    print(emsg)
                else:
                    print(emsg)
                    raise Exception(emsg)

        for w in ['warning', 'Warning', 'warnings', 'Warnings']:
            # warning can be null (python None).
            if w in rjson and rjson[w]:
                H2O.verboseprint(dump_json(rjson))
                print('rjson %s in %s: %s' % (w, inspect.stack()[1][3], rjson[w]))


        # Allow the caller to check things like __http_request.status_code.
        # The response object is not JSON-serializable, so we capture the fields we want here:
        response = {}
        # response['headers'] = r.headers
        response['url'] = r.url
        response['status_code'] = r.status_code
        response['text'] = r.text
        rjson['__http_response'] = response

        return rjson
示例#2
0
    def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
                          cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, **kwargs):
        H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
        # if url param is used, use it as full url. otherwise crate from the jsonRequest
        if fullUrl:
            url = fullUrl
        else:
            url = self.__url(jsonRequest)

        # remove any params that are 'None'
        # need to copy dictionary, since can't delete while iterating
        if params is not None:
            params_serialized = params.copy()
            for k in params_serialized:
                if params_serialized[k] is None:
                    del params[k]
            paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
        else:
            paramsStr = ''

        # The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
        # NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
        # NOTE: we currently don't need to do this for GET, so that's not implemented.
        if postData is not None:
            munged_postData = {}
            for k, v in postData.iteritems():
                if type(v) is list:
                    if len(v) == 0:
                        munged_postData[k] = '[]'
                    else:
                        first = True
                        array_str = '['
                        for val in v:
                            if not first: array_str += ', '

                            if val is None:
                                array_str += 'null'
                            elif isinstance(val, basestring):
                                array_str += "\"" + str(val) + "\""
                            else:
                                array_str += str(val)
                            first  = False
                        array_str += ']'
                        munged_postData[k] = array_str
                else:
                    # not list:
                    munged_postData[k] = v
        else:  
            # None
            munged_postData = postData

        if extraComment:
            log('Start ' + url + paramsStr, comment=extraComment)
        else:
            log('Start ' + url + paramsStr)

        log_rest("")
        log_rest("----------------------------------------------------------------------\n")
        if extraComment:
            log_rest("# Extra comment info about this request: " + extraComment)
        if cmd == 'get':
            log_rest("GET")
        else:
            log_rest("POST")
        log_rest(url + paramsStr)

        # file get passed thru kwargs here
        try:
            if 'post' == cmd:
                # NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
                # This is temporary.
                # 
                # This following does application/json (aka, posting JSON in the body):
                # r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
                # 
                # This does form-encoded, which doesn't allow POST of nested structures
                r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
            elif 'delete' == cmd:
                r = requests.delete(url, timeout=timeout, params=params, **kwargs)                
            elif 'get' == cmd:
                r = requests.get(url, timeout=timeout, params=params, **kwargs)
            else:
                raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)

        except Exception, e:
            # rethrow the exception after we've checked for stack trace from h2o
            # out of memory errors maybe don't show up right away? so we should wait for h2o
            # to get it out to h2o stdout. We don't want to rely on cloud teardown to check
            # because there's no delay, and we don't want to delay all cloud teardowns by waiting.
            # (this is new/experimental)
            exc_info = sys.exc_info()
            # use this to ignore the initial connection errors during build cloud when h2o is coming up
            if not noExtraErrorCheck: 
                h2p.red_print(
                    "ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
                time.sleep(2)
                H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
            log_rest("")
            log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
            raise exc_info[1], None, exc_info[2]

            H2O.verboseprint("r: " + repr(r))
示例#3
0
    def __do_json_request(self, jsonRequest=None, fullUrl=None, timeout=10, params=None, postData=None, returnFast=False,
                          cmd='get', extraComment=None, ignoreH2oError=False, noExtraErrorCheck=False, raiseIfNon200=True, suppressErrorMsg=False, **kwargs):
        H2O.verboseprint("__do_json_request, timeout: " + str(timeout))
        # if url param is used, use it as full url. otherwise crate from the jsonRequest
        if fullUrl:
            url = fullUrl
        else:
            url = self.__url(jsonRequest)

        # remove any params that are 'None'
        # need to copy dictionary, since can't delete while iterating
        if params is not None:
            params_serialized = params.copy()
            for k in params_serialized:
                if params_serialized[k] is None:
                    del params[k]
            paramsStr = '?' + '&'.join(['%s=%s' % (k, v) for (k, v) in params.items()])
        else:
            paramsStr = ''

        # The requests package takes array parameters and explodes them: ['f00', 'b4r'] becomes "f00,b4r".
        # NOTE: this handles 1D arrays only; if we need ND this needs to be recursive.
        # NOTE: we currently don't need to do this for GET, so that's not implemented.
        if postData is not None:
            munged_postData = {}
            for k, v in postData.iteritems():
                if type(v) is list:
                    if len(v) == 0:
                        munged_postData[k] = '[]'
                    else:
                        first = True
                        array_str = '['
                        for val in v:
                            if not first: array_str += ', '

                            if val is None:
                                array_str += 'null'
                            elif isinstance(val, basestring):
                                array_str += "\"" + str(val) + "\""
                            else:
                                array_str += str(val)
                            first  = False
                        array_str += ']'
                        munged_postData[k] = array_str
                elif type(v) is dict:
                    if len(v) == 0:
                        munged_postData[k] = '{}'
                    else:
                        first = True
                        map_str = '{'
                        for key, val in v.iteritems():
                            if not first: map_str += ', '

                            if val is None:
                                map_str += "\"" + key + "\"" + ': null'
                            elif isinstance(val, basestring):
                                map_str += "\"" + str(key) + "\"" + ":" + "\"" + str(val) + "\""
                            else:
                                map_str += "\"" + key + "\"" + ':' + str(val)
                            first  = False
                        map_str += '}'
                        munged_postData[k] = map_str

                else:
                    # not list:
                    munged_postData[k] = v
        else:
            # None
            munged_postData = postData

        # print("munged_postData: " + repr(munged_postData))

        if extraComment:
            log('Start ' + url + paramsStr, comment=extraComment)
        else:
            log('Start ' + url + paramsStr)

        log_rest("")
        log_rest("----------------------------------------------------------------------\n")
        if extraComment:
            log_rest("# Extra comment info about this request: " + extraComment)
        if cmd == 'get':
            log_rest("GET")
        else:
            log_rest("POST")
        log_rest(url + paramsStr)

        # file get passed thru kwargs here
        try:
            if 'post' == cmd:
                # NOTE == cmd: for now, since we don't have deserialization from JSON in h2o-dev, we use form-encoded POST.
                # This is temporary.
                #
                # This following does application/json (aka, posting JSON in the body):
                # r = requests.post(url, timeout=timeout, params=params, data=json.dumps(munged_postData), **kwargs)
                #
                # This does form-encoded, which doesn't allow POST of nested structures
                r = requests.post(url, timeout=timeout, params=params, data=munged_postData, **kwargs)
            elif 'delete' == cmd:
                r = requests.delete(url, timeout=timeout, params=params, **kwargs)
            elif 'get' == cmd:
                r = requests.get(url, timeout=timeout, params=params, **kwargs)
            else:
                raise ValueError("Unknown HTTP command (expected 'get', 'post' or 'delete'): " + cmd)

        except Exception as e:
            # rethrow the exception after we've checked for stack trace from h2o
            # out of memory errors maybe don't show up right away? so we should wait for h2o
            # to get it out to h2o stdout. We don't want to rely on cloud teardown to check
            # because there's no delay, and we don't want to delay all cloud teardowns by waiting.
            # (this is new/experimental)
            exc_info = sys.exc_info()
            # use this to ignore the initial connection errors during build cloud when h2o is coming up
            if not noExtraErrorCheck:
                h2p.red_print(
                    "ERROR: got exception on %s to h2o. \nGoing to check sandbox, then rethrow.." % (url + paramsStr))
                time.sleep(2)
                H2O.check_sandbox_for_errors(python_test_name=H2O.python_test_name);
            log_rest("")
            log_rest("EXCEPTION CAUGHT DOING REQUEST: " + str(e.message))
            raise (exc_info[1], None, exc_info[2])

            H2O.verboseprint("r: " + repr(r))

        if 200 != r.status_code:
            pp = pprint.PrettyPrinter(indent=4)
            msg = "JSON call returned non-200 status: " + url

            json = r.json()
            if None != json and 'dev_msg' in json:
                msg += "\ndev_msg: "
                msg += str(json['dev_msg'])
            msg += "\nr.status_code: " + str(r.status_code)
            msg += "\nr.headers: " + repr(r.headers)
            if None == json:
                msg += '\nERROR: the error output from H2O is not JSON!'
                msg += "\nr.text: " + r.text
            else:
                msg += "\nr.json: "
                msg += pp.pformat(json)

            if raiseIfNon200:
                pass  # we'll pass msg up with the exception
            elif not suppressErrorMsg:
                print(msg)
            log_rest(msg)

        log_rest("")
        try:
            if r is None:
                log_rest("r is None")
            else:
                log_rest("HTTP status code: " + str(r.status_code))
                # The following accesses to r.text were taking most of the runtime:
                log_text = False
                if log_text:
                    if hasattr(r, 'text'):
                        if r.text is None:
                            log_rest("r.text is None")
                        else:
                            log_rest(r.text)
                    else:
                        log_rest("r does not have attr text")
        except Exception as e:
            # Paranoid exception catch.
            # Ignore logging exceptions in the case that the above error checking isn't sufficient.
            print("Caught exception from result logging: ", e, "; result: ", repr(r))

        # fatal if no response
        if raiseIfNon200 and not r:
            raise Exception("Maybe bad url? no r in __do_json_request in %s:" % inspect.stack()[1][3] + "\n\n" + msg)

        # this is used to open a browser on results, or to redo the operation in the browser
        # we don't' have that may urls flying around, so let's keep them all
        H2O.json_url_history.append(r.url)
        # if r.json():
        #     raise Exception("Maybe bad url? no r.json in __do_json_request in %s:" % inspect.stack()[1][3])

        rjson = None
        if returnFast:
            return
        try:
            rjson = r.json()
        except:
            print(h2o_test_utils.dump_json(r.text))
            if not isinstance(r, (list, dict)):
                raise Exception("h2o json responses should always be lists or dicts, see previous for text")

            raise Exception("Could not decode any json from the request.")

        # TODO
        # TODO
        # TODO
        # TODO: we should really only look in the response object.  This check
        # prevents us from having a field called "error" (e.g., for a scoring result).
        for e in ['error', 'Error', 'errors', 'Errors']:
            # error can be null (python None). This happens in exec2
            if e in rjson and rjson[e]:
                H2O.verboseprint("rjson:" + h2o_test_utils.dump_json(rjson))
                emsg = 'rjson %s in %s: %s' % (e, inspect.stack()[1][3], rjson[e])
                if ignoreH2oError:
                    # well, we print it..so not totally ignore. test can look at rjson returned
                    print(emsg)
                else:
                    print(emsg)
                    raise Exception(emsg)

        for w in ['warning', 'Warning', 'warnings', 'Warnings']:
            # warning can be null (python None).
            if w in rjson and rjson[w]:
                H2O.verboseprint(dump_json(rjson))
                print('rjson %s in %s: %s' % (w, inspect.stack()[1][3], rjson[w]))


        # Allow the caller to check things like __http_request.status_code.
        # The response object is not JSON-serializable, so we capture the fields we want here:
        response = {}
        # response['headers'] = r.headers
        response['url'] = r.url
        response['status_code'] = r.status_code
        response['text'] = r.text
        rjson['__http_response'] = response

        return rjson