Mailing-List: contact user-help@avro.apache.org; run by ezmlm
Precedence: bulk
Reply-To: user@avro.apache.org
MIME-Version: 1.0
In-Reply-To: <CAEHYH9_DnuUGu8B6uUsxxrfO1kcUxPcqp77tnegKuxVKi67bhA@mail.gmail.com>
References: <CAEHYH987GZjuERn4Zy4sh4H2zfAbTJmCrkD5Tcjugi4UhmYZTA@mail.gmail.com>
 <CADNPUY58O5iL=DivGCD2aDBvXXfw5FhnqAoHFSa8CrY1KJDKGQ@mail.gmail.com>
 <CAEHYH9-UgY-tSh1QR5MToyy=Vijw3ns=JWS3WNJAqiV6jQQxfg@mail.gmail.com>
 <CADNPUY7h_oJwngqdbe3G0WpTiJVjvZsc0GMtSg=ZPPqAuevzKA@mail.gmail.com>
 <CAEHYH98gxAje9j2eaygKTECRGmoVzeXWnK9iSxHWSJc63kMotg@mail.gmail.com>
 <CADNPUY71NyZceMphnkv1-uRbivKps0Ad4KF5Je4DXS6wrHQRZQ@mail.gmail.com> <CAEHYH9_DnuUGu8B6uUsxxrfO1kcUxPcqp77tnegKuxVKi67bhA@mail.gmail.com>
From: Dan Schmitt <dan.schmitt@gmail.com>
Date: Mon, 27 Nov 2017 12:47:10 -0500
Message-ID: <CADNPUY5urR90FKHQXi_4mgB-eFjL55AT++xkEUGZkuknWjmK0A@mail.gmail.com>
Subject: Re: how to do non-optional field in nested object?
To: user@avro.apache.org
Content-Type: text/plain; charset="UTF-8"
archived-at: Mon, 27 Nov 2017 17:47:23 -0000

I'm using the C++ api to do validation outside what I can specify in
the schema spec.

On Mon, Nov 27, 2017 at 11:12 AM, Martin Mucha <alfonz19@gmail.com> wrote:
> Following schema:
> {
>   "name" : "ARecord",
>   "type" : "record",
>   "namespace" : "AAA",
>   "fields" : [
>     {"name": "A", "type": "string" }
>   ]
> }
>
>
> does not validate json
>
> {}
>
> as valid one. It's invalid. A is required, as would be expected by me.
>
> ---
> I'm getting lost here.
> What is the resolution? I kinda wanted to use avro schema to validate
> described JSON, and now based on what you said IIUC validation in "xschema
> style": "1 of this, followed by 2 of that" is not possible with avro schema.
> Correct?
>
>
> Btw. what do you use to validate JSON using avro? I used avro-utils.jar
> executed from command line, which proved as incapable of deserializing &
> validating optional fields if they are set (if optional field is set, I have
> to pass value in json like: {"string":"value"}). So now I'm using for
> testing purposes actual flow through NIFI, which is extremely cumbersome.
>
> 2017-11-27 16:19 GMT+01:00 Dan Schmitt <dan.schmitt@gmail.com>:
>>
>> The top level object in all the examples is a record (of which you can
>> have 0 or more.)
>>
>> So, right now, even the top level is failing the spec:
>>
>> IV) valid (0 ARecords):
>> { }
>>
>> V) valid (2 ARecords):
>> {
>>   "id": "...",
>>   "B": {
>>     "C": "..."
>>   }
>> } ,
>> "id": "...",
>>   "B": {
>>     "C": "..."
>>   }
>> }
>>
>> On Mon, Nov 27, 2017 at 9:47 AM, Martin Mucha <alfonz19@gmail.com> wrote:
>> > I don't understand where "or more" is comming from.
>>
>> Because of the use of nested records.  Anywhere you put a record you can
>> have 0,
>> or more than 1.  I don't know of a pure schema way to enforce only
>> 1/optional.
>>
>> If you are doing something with any of the APIs you could add a validation
>> step
>> that says "must have only 1 ARecord, and must have only 1 BRecord" after
>> you
>> read the data and throw an error for the 0 or 1+ situations, but you'd
>> need to write
>> some code somewhere with one of the APIs and build your own validator.
>
>