'?' always replaced as parameter (even in string literals and comments)

[hochhaus]

Thanks for all of the work @denisenkom and @dimdin. It is exciting how quickly go-mssqldb is maturing. I am a bit concerned with change 0a8a8cf that uses '?' for parameters. I'm interested in both your thoughts about the best way to proceed.

It looks like if go-mssqldb uses '?' the library will need to parse the query and replace '?' with @px (where X is the integer field number). However, parsing the query in the client side library seems error prone to get all of the cases correct (eg: string literals, comments, etc). We could definitely improve the implementation of parseParams() to handle these cases, however it seems like it might be better to avoid parsing the query all together client side (if possible).

The plus side is that '?' does seem like the standard character to use for MS SQL so I like the change from this perspective.

Does anyone know how other mssql client libraries are doing this? I know nothing about TDS so I'm not sure of the limitations we are facing at the protocol level.

Thoughts?

[dimdin]

I shall extend the parseParams to scan the SQL statement and replace the ? only in the appropriate state i.e. not in comments, strings and identifiers.

[denisenkom]

Underlying protocol only supports "at" prefixed parameter names, support for question marks is supported by client library, freetds does exactly this.
BTW it doesn't have to be question marks, it might be something more appropriate to given platform, for example pytds supports format and pyformat http://legacy.python.org/dev/peps/pep-0249/#paramstyle

[hochhaus]

Thanks for the comments and patch! This makes sense.

[dimdin]

It makes sense to have positional positional parameters, :1, :2 like Oracle or $1, $2 like PostgreSQL.
The patch can be extended to support them.