MSSQL: Column name mapping breaks for certain naming patterns

[amitdeshmukh]

Description

When using GraphJin with MSSQL (SQL Server), certain column names are incorrectly mapped in the generated SQL. GraphJin appears to apply a camelCase transformation that strips underscores from some column names, producing SQL column names that don't match the actual database schema.

Environment

• GraphJin version: 3.16.2
• Database: Microsoft SQL Server 2016 (named instance)
• OS: macOS (darwin)

Reproduction

Given an MSSQL table Theatre_Master with columns like Is_Active, Is_DCI, property_id, Equipment_Type:

Works correctly

{ theatre_master(limit: 3) { theater_id theater_name city ckt_id region_id } }

These columns map correctly:

• theater_id → Theater_Id ✅
• theater_name → Theater_Name ✅
• ckt_id → Ckt_Id ✅
• region_id → Region_ID ✅

Broken columns

{ theatre_master(limit: 3) { is_active is_dci property_id equipment_type } }

GraphQL field	GraphJin generates	Actual DB column	Error
is_active	IsActive	Is_Active	mssql: Invalid column name 'IsActive'
is_dci	IsDCI	Is_DCI	mssql: Invalid column name 'IsDCI'
property_id	propertyId	property_id	mssql: Invalid column name 'propertyId'
equipment_type	(not found)	Equipment_Type	field 'equipment_type' is not a column or a function

The same pattern occurs across multiple tables — any column with is_ prefix or a lowercase word_word pattern gets incorrectly camelCased.

Expected behavior

GraphJin should use the exact column names as introspected from INFORMATION_SCHEMA.COLUMNS (or sys.columns) when generating SQL for MSSQL. The camelCase transformation should not be applied to MSSQL column names since MSSQL schemas commonly use mixed naming conventions (PascalCase_With_Underscores, lowercase_with_underscores, Is_PascalCase).

Config

Using auto-discovery (tables: []) with no explicit column configuration:

databases:
  default:
    type: mssql
    host: 10.25.5.25
    port: 5205
    dbname: VMPL_DATA
    user: myuser
    password: mypass

Workaround

Currently using node-mssql for queries involving affected columns. Would prefer to use GraphJin for all reads.