Issue where clause FOR JSON AUTO has generated the incomplete answer [duplicate] - sql-server

Question

This question already has answers here:
FOR JSON PATH results in SSMS truncated to 2033 characters
(10 answers)
SQL Server json truncated (even when using NVARCHAR(max) )
(10 answers)
Closed 9 months ago.
Getting JSON from SQL Server is great, but I ran into a problem.
Example. I have a LithologySamples table with a very basic structure:
[Id] [uniqueidentifier],
[Depth1] [real],
[Depth2] [real],
[RockId] [nvarchar](8),
In the database there are more or less 600 records of this table. I want to generate a JSON to transport data to another database, so I use FOR JSON AUTO. Which has worked perfectly with other tables with less records. But in this case I see that the response is generated incomplete. It has me baffled. I noticed when examining the output:
[{
"Id": "77769039-B2B7-E511-8279-DC85DEFBF2B6",
"Depth1": 4.2000000e+001,
"Depth2": 5.8000000e+001,
"RockId": "MIC SST"
}, {
"Id": "78769039-B2B7-E511-8279-DC85DEFBF2B6",
"Depth1": 5.8000000e+001,
"Depth2": 6.3000000e+001,
"RockId": "CGL"
}, {
"Id": "79769039-B2B7-E511-8279-DC85DEFBF2B6",
"Depth1": 6.3000000e+001,
"Depth2": 8.3000000e+001,
"RockId": "MIC SST"
}, {
// ... OK, continue fine, but it breaks off towards the end:
}, {
"Id": "85769039-B2B7-E511-8279-DC85DEFBF2B6",
"Depth1": 2.0500000e+002,
"Depth2": 2.1500000e+002,
"RockId": "MIC SST"
}, {
"Id": "86769039-
// inexplicably it cuts here !?
I've searched and I can't find any options for the answer to come out complete.
The SQL query is as follows:
SELECT*FROM LithologySamples FOR JSON AUTO;
AUTO or PATH are the same result
Does anyone know what I should do so that the statement generates the JSON of the entire table?

Answer 1

But in this case I see that the response is generated incomplete.
If you are checking this in SSMS, it truncates text in various ways depending on the output method you're using (PRINT, SELECT, results to text/grid). The string is complete, it's just the output that has been mangled.
One way to validate that the string is in fact complete is to:
SELECT * INTO #foo FROM
(SELECT * FROM LithologySamples FOR JSON AUTO) x(y);
Then checking LEN(y), DATALENGTH(y), RIGHT(y , 50) (see example db<>fiddle), or selecting from that table using CONVERT(xml (see this article for more info).
In your case it seems the problem is coming from how C# is consuming the output. If the consumer is treating the JSON as multiple rows, then assigning a variable there will ultimately assign one arbitrary row of <= 2033 characters, not the whole value. I talked about this briefly back in 2015. Let's say you are using reader[0] or similar to test:
CREATE TABLE dbo.Samples
(
[Id] [uniqueidentifier] NOT NULL DEFAULT NEWID(),
[Depth1] [real] NOT NULL DEFAULT 5,
[Depth2] [real] NOT NULL DEFAULT 5,
[RockId] [nvarchar](8)
);
INSERT dbo.Samples(RockId) SELECT TOP (100) LEFT(name, 8) FROM sys.all_columns;
-- pretend this is your C# reader:
SELECT * FROM dbo.Samples FOR JSON AUTO;
-- reader[0] here would be something like this:
-- [{"Id":"054EC9A2-760B-4EBA-BF06-...,"RockId":"ser
-- which is the first 2,033 characters
SELECT LEN('[{"Id":"054EC9A2-760B-4EBA-BF06-..."RockId":"ser')
-- instead, since you want C# to assign a scalar,
-- assign output to a scalar first:
DECLARE #json nvarchar(max) = (SELECT * FROM dbo.Samples FOR JSON AUTO);
SELECT json = #json;
-- now reader[0] will be the whole thing
Example db<>fiddle
The 2033 comes from the same place it comes from for XML (since SQL Server's JSON implementation is just a pretty wrapper under existing underlying XML functionality), as Charlie points out Martin explained here:
SELECT FOR XML AUTO and return datatypes