Error when attempting to read resultset with VARCHAR column

[Adam Wilson]

Using the latest ODBC 17 Driver from Microsoft, the following error occurs when attempting to read a query ResultSet that contains a column of type VARCHAR. (n) or (MAX) does not make a difference.

SQL QUERY:
SELECT ss.[schema_id], [Name] = CONVERT(VARCHAR(8000), ss.[name]) FROM [sys].[schemas] AS ss WHERE ss.[name] <> 'sys' AND ss.[name] <> 'guest' AND ss.[name] <> 'INFORMATION_SCHEMA' AND ss.[name] NOT LIKE 'db[_]%'

STACK TRACE:
2023-04-24T01:33:03.004 [info] C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d:428:this Driver=ODBC Driver 17 for SQL Server;Server=tcp:192.168.0.6;Uid=sa;Pwd=;Database=SERVICEINSIGHT-DEV
2023-04-24T01:33:03.209 [error] C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d:110:check odbc.sql.SQLGetData(1DDFB9974A0, 2, 1, null, 5, null) : SQL_ERROR

object.Exception@C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(850):
HY009:1:0 [Microsoft][ODBC Driver Manager] Invalid argument value

0x00007FF7572CAC87 in d_throwc
0x00007FF7571383B4 in ddbc.drivers.odbcddbc.extractError at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(151)
0x00007FF75713C4D8 in ddbc.drivers.odbcddbc.check!(SQLGetData, "C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d", 850LU).check at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(119)
0x00007FF75711EE0E in ddbc.drivers.odbcddbc.ODBCStatement.checkstmt!(SQLGetData, "C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d", 850LU).checkstmt at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(587)
0x00007FF7570F83E2 in ddbc.drivers.odbcddbc.ODBCStatement.ColumnInfo.readValue!(char[]).readValue at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(850)
0x00007FF7570DE2DB in ddbc.drivers.odbcddbc.ODBCStatement.ColumnInfo.readValueAsVariant at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(932)
0x00007FF7570DE058 in ddbc.drivers.odbcddbc.ODBCStatement.ColumnInfo.read at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(809)
0x00007FF75711E767 in ddbc.drivers.odbcddbc.ODBCStatement.fetch.__lambda2 at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(768)
0x00007FF7570F7B97 in ddbc.drivers.odbcddbc.ODBCStatement.fetch.each!(ddbc.drivers.odbcddbc.ODBCStatement.ColumnInfo[]).each at C:\D\dmd2\windows\bin64....\src\phobos\std\algorithm\iteration.d(1000)
0x00007FF7570DDE8A in ddbc.drivers.odbcddbc.ODBCStatement.fetch at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(768)
0x00007FF7570E1CE5 in ddbc.drivers.odbcddbc.ODBCResultSet.next at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(1416)
0x00007FF7573A6C47 in sqlbinding.mssql.schemareader.ReadSchemata at C:\Users\AdamWilson\Projects\EllipticBit\SqlBinding\source\mssql\schemareader.d(30)

[Samael]

that's odd. The CI does test with SQL Server using the msodbcsql18 driver and SQL Server 2022 and the table structure for the test certainly uses VARCHAR:

ddbc/example/source/testddbc.d

Lines 328 to 360 in 266d935

 		case "odbc": 
 			stmt.executeUpdate("DROP TABLE IF EXISTS [ddbct1]"); 
 			stmt.executeUpdate("CREATE TABLE ddbct1 ( 
 				[id] INT NOT NULL IDENTITY(1,1) PRIMARY KEY, 
 				[name] VARCHAR(250), 
 				[comment] VARCHAR(max), 
 				[ts] DATETIME 
 				)"); 
  
 			stmt.executeUpdate("INSERT INTO [ddbct1] ([name], [comment], [ts])  
 								VALUES  
 									('name1', 'comment for line 1', CURRENT_TIMESTAMP),  
 									('name2','comment for line 2 - can be very long', CURRENT_TIMESTAMP)"); 
 			 
 			stmt.executeUpdate("DROP TABLE IF EXISTS [employee]"); 
 			stmt.executeUpdate("CREATE TABLE [employee] ( 
 				[id] INT NOT NULL IDENTITY(1,1) PRIMARY KEY, 
 				[name] VARCHAR(255) NOT NULL, 
 				[flags] int null, 
 				[dob] DATE, 
 				[created] DATETIME default CURRENT_TIMESTAMP, 
 				[updated] DATETIMEOFFSET default CURRENT_TIMESTAMP 
 				)"); 
  
 			stmt.executeUpdate(`INSERT INTO [employee] ([name], [flags], [dob], [created], [updated]) 
 								VALUES 
 									('John', 5, '1976-04-18', '2017-11-23 20:45', '2010-12-30 00:00:00'), 
 									('Andrei', 2, '1977-09-11', '2018-02-28 13:45', '2010-12-30 12:10:12'), 
 									('Walter', 2, '1986-03-21', '2018-03-08 10:30', '2010-12-30 12:10:04.100'), 
 									('Rikki', 3, '1979-05-24', '2018-06-13 11:45', '2010-12-30 12:10:58'), 
 									('Iain', 0, '1971-11-12', '2018-11-09 09:33', '2010-12-30 12:10:01'), 
 									('Robert', 1, '1966-03-19', CURRENT_TIMESTAMP, CURRENT_TIMESTAMP)`); 
             break; 

Does the SQL run correctly when run directly against the database?

[Adam Wilson]

Yes, it works in SSMS. I am using msodbcsql17 and SQL Server 2019.

After further review it appears that the test only attempts to use a '*' SELECT list and doesn't test named select lists.

When I use a * selector I get the following error:

std.variant.VariantException@std\variant.d(1832): Variant: attempting to use incompatible types wchar[] and immutable(char)[]

0x00007FF7DA93AC87 in d_throwc
0x00007FF7DA76B73A in std.variant.VariantN!20LU.VariantN.get!string.get at C:\D\dmd2\windows\bin64....\src\phobos\std\variant.d(862)
0x00007FF7DA7529BB in ddbc.drivers.odbcddbc.ODBCResultSet.getString at C:\Users\AdamWilson\AppData\Local\dub\packages\ddbc-0.5.7\ddbc\source\ddbc\drivers\odbcddbc.d(1573)
0x00007FF7DAA16C9A in sqlbinding.mssql.schemareader.ReadSchemata at C:\Users\AdamWilson\Projects\EllipticBit\SqlBinding\source\mssql\schemareader.d(34)
0x00007FF7DA6A7A8A in D main at C:\Users\AdamWilson\Projects\EllipticBit\SqlBinding\source\app.d(92)*

I assume this error is because every string in the information schema is NVARCHAR and DDBC doesn't support NVARCHAR yet?
The trouble is that because named select lists don't work, I can't do a CONVERT(VARCHAR) to get something that it can read.

[Samael]

yeah #73 was created to add support for NVARCHAR but it didn't get implemented.

There are other select statements that are tested. it does:

• SELECT * FROM ddbct1
• SELECT id,comment FROM ddbct1 WHERE id = 2
• SELECT id, comment, ts FROM ddbct1 ORDER BY id DESC
• SELECT id, name, comment, ts FROM ddbct1 WHERE ts IS NULL
• SELECT id, name name_alias, comment, ts FROM ddbct1 WHERE id >= ?

as well as select statements that are generated via pod support. The error you're seeing indicates that the issue is with the variable type being used in the D code. Are you retrieving the data as a ResultSet then calling getString or using a pod?

[Adam Wilson]

I'm using ResultSet.getString(n). It's actually code that I ported from C#/ADO.NET which has a very similar model to DDBC/JDBC ResultSets.

So I did some more digging, and here is what I came up with. The original error posted disappears when my SELECT list columns are in the same order as the original table column order. When I use the select list to reorder it, I get that HY009 error.

[Samael]

that's interesting, especially as those tests differ in columns used. perhaps the alias for the table name makes a difference. It may be worth adding test cases for variations of the SELECT id, name, comment, ts FROM ddbct1 WHERE ts IS NULL statement such as:

• SELECT id, ts, comment, name FROM ddbct1 WHERE ts IS NULL
• SELECT t1.id, t1.name, t1.comment, t1.ts FROM ddbct1 t1 WHERE ts IS NULL

[Adam Wilson]

Those look good. I can work around this by reordering my select lists, but I am puzzled as to why DDBC/ODBC even cares because I'm using column indexes for my getters.