Success! I had to fix another trivial bug in queryserver.py to allow the *-Dphoenix.queryserver.**serialization=JSON* to get passed to the Java command.
Then I ran into a pythondb error very much like the one described at https://bitbucket.org/lalinsky/python-phoenixdb/issues/1/error-on-avatica-150 . So, I looked at the committed bug fix <https://bitbucket.org/lalinsky/python-phoenixdb/commits/65ed3c64d830> and guess that I needed to wrap everything up in a openConnection / closeConnection pair. Also, the verbose log from shell.py --debug , when compared to the above verbose log, looked like the JSON was still being sent in the header instead of the body. I'm guessing the format is DEBUG:phoenixdb.avatica:POST / *<body> <header>* These two incidents leads me to guess that root cause of my issue with pythondb is that it was not correctly recognizing the Avatica version. Anyway, I got upserts working remotely with curl, with the request JSON in the body. I hope that having the request in the body will now allow me to make these calls in Javascript. (See my other email topic for details on *that*.) Thanks again, Lukas. You can email me directly if you want me to help troubleshoot pythondb for https://dist.apache.org/repos/dist/dev/phoenix/phoenix-4.7.0-HBase-0.98-rc1/ . Bye, Steve On Fri, Feb 5, 2016 at 3:07 PM, Steve Terrell <[email protected]> wrote: > Thanks, Lukas. Half the battle is won, now. With your help I was able to > see the JSON used to perform the upsert. Looks like my problem was that I > was not doing any "connectionSync" calls. I had played around with the > "createStatement" before my upsert SQL, but I may not have stumbled across > the proper sequence. > > It would have been nice if somewhere there were Avatica docs to let people > know what the sequence of calls must be. For example: first you do > connectionSync, second you do createStatement, then you > do prepareAndExecute, etc. > > I'll try your tip for starting the 4.7 server in JSON mode sometime soon. > > On Fri, Feb 5, 2016 at 1:59 PM, Lukáš Lalinský <[email protected]> wrote: > >> On Fri, Feb 5, 2016 at 8:46 PM, Steve Terrell <[email protected]> >> wrote: >>> >>> When I tried to send a "createStatement" via curl and via Lukas's >>> phoenixdb, I got these error's respectively: >>> >>> <h2>HTTP ERROR: 500</h2> >>> <p>Problem accessing /. Reason: >>> <pre> Cannot find parser for </pre></p> >>> >>> >>> File "/usr/local/lib/python2.7/site-packages/phoenixdb/avatica.py", >>> line 100, in parse_error_page >>> raise errors.InternalError(message) >>> phoenixdb.errors.InternalError: ('Cannot find parser for', None, None, >>> None) >>> >>> >>> Too bad - I think I need 4.7 to pass JSON in the HTTP request body. >>> >> >> 4.7 uses Protobuf3 serialization by default, but you can still make it >> use JSON: >> >> ./bin/queryserver.py start -Dphoenix.queryserver.serialization=JSON >> >> >>> However, I *was* able to get phoenixdb working with my Phoenix 4.6, >>> both selects and upserts, both locally and remotely. So, I'd like to take >>> Lukas's advice and log what is being sent so I can replicate it outside of >>> Python. >>> >>> Does anyone know: >>> >>> - How to log the headers and body of all requests >>> >>> You should be able to run this and see the request/response details. >> >> ./examples/shell.py --debug http://localhost:8765/ >>> >>> >>> - Where the Avatica documentation is for how to make queries for >>> different versions? As I linked in my original email, all I could find >>> was >>> syntax of the latest JSON, bit no docs on headers, post body, how to form >>> the HTTP request, etc. >>> >>> There is no documentation for the older versions of the protocol, as far >> as I know. I based my library on the Java code in calcite. >> >> Lukas >> >> >
