[MSSQL]DataTypes.STRING should be Varchar or NVarchar ?

[p2227]

version:3.19.3

http://docs.sequelizejs.com/en/latest/api/datatypes/

according to the docs, DataTypes.STRING should be map to Varchar. but i got nvarchar in mssql ?

[janmeier]

sequelize/lib/dialects/mssql/data-types.js

Line 48 in 16b39c6

return 'NVARCHAR(' + this._length + ')';

[komeilshahmoradi]

version 4.42.0
I still have this problem
please, how to solve this problem ?

[durango]

@komeilshahmoradi check out this issue #8533 might help guide you on how to extend data types to your liking. I am not sure why SequelizeJS decided to opt for nvarchar but it is, in my opinion, incorrect. I'll ask the team and see why this decision was made and what are our options for the future.

Alternatively, if you don't mind using varchar(255) you can use ENUM but this isn't really ideal

sequelize/lib/dialects/mssql/data-types.js

Line 175 in 16b39c6

ENUM.prototype.toSql = function() {

Reopening because it is an issue

[bjagadeesan]

version: 4.42.0

I have the same problem. I need to use varchar instead of nvarchar. I execute raw queries through sequelize. The replacements always converts to nvarchar and it is a performance bottleneck for my operation. The regular query with varchar runs in under a second and the one with nvarchar runs anywhere from 30 to 45 seconds.

I tried overriding the tosql() method, but I couldn't get the extended version to be used in the program. Any help on that is much appreciated.

Below is the code .

import {DataTypes} from "sequelize";

DataTypes.STRING.prototype.toSql = function() {
	if (!this._binary) {
		return "VARCHAR(" + this._length + ")";
	} else{
		return "BINARY(" + this._length + ")";
	}
};

export {DataTypes};

But ideally, adding an option to choose between nvarchar and varchar will be much helpful.

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. If this is still an issue, just leave a comment 🙂

[SheetalGoel]

I am using "sequelize": "^5.21.5"
And still this issue causing a major impact on our operations.

[nareshy-quisitive]

Any update on this issue? Its impacting application performace.

[WikiRik]

@ephys I think you can answer this

[nareshy-quisitive]

Issue is

const searchResults = await models.file.findAll({
        where: {
          path: {
            [Op.like]: `%${queryStr}%`
          }
        },
        attributes: ['path', 'uuid']
      })

Above code executing following query
select [uuid],[path] from [omn].[file] where [path] like N'%abc%'
which is taking around 7sec to complete the query. But if I execute the same query by removing 'N' from the query I am getting the same response within 1sec.

Note: path is varchar data type in MSSQL DataBase

[ephys]

I don't want to change what DataTypes.STRING produces, even in a SemVer MAJOR release, because it would make migrating too difficult. But I would like to deprecate DataTypes.STRING, see #14259

In the meantime, there is a workaround you can use that we just documented. You can use VARCHAR like this in mssql:

As for generating 'string' instead of N'string', it will need to wait until we have DataTypes.VARCHAR & DataTypes.VARCHAR.N so we can generate the appropriate string based on the DataType. This part will be tracked in #9107

[nareshy-quisitive]

@ephys suggestions are not working. We are using the 5.22.4 version. Do we need to upgrade this to the latest version?

[ephys]

I did a quick test in a SSCCE running with mssql and the following code:

class User extends Model {}

User.init({
  firstName: 'VARCHAR(50)',
}, {
  sequelize,
});

await sequelize.sync({ force: true });

Which resulted in the following SQL:

IF OBJECT_ID('[Users]', 'U') IS NOT NULL DROP TABLE [Users];

IF OBJECT_ID('[Users]', 'U') IS NULL CREATE TABLE [Users] ([id] INTEGER NOT NULL IDENTITY(1,1) , [firstName] VARCHAR(50) NULL, [createdAt] DATETIMEOFFSET NOT NULL, [updatedAt] DATETIMEOFFSET NOT NULL, PRIMARY KEY ([id]));

EXEC sys.sp_helpindex @objname = N'[Users]';

I was running it using Sequelize 6, it may be the case that this is not supported in Sequelize 5. There is a small number of breaking changes in Sequelize 6, I definitely recommend upgrading: https://sequelize.org/master/manual/upgrade-to-v6.html

[nareshy-quisitive]

I have tested this with the latest version 6.18.0. still facing the same issue
here is my test model class

const config = require('../config/index')
const _ = require('lodash')
const sequelize = new Sequelize(config.get('database'), config.get('username'), config.get('password'), {
  dialect: config.get('dialect'),
  host: config.get('host'),
  logging: _.get(config, 'logging', console.log),
  benchmark: true,
  port: 1433,
  dialectOptions: {
    options: {
      encrypt: true,
      trustServerCertificate: true
    },
  }
})
class File extends Model { }

File.init({
  uuid: {
    allowNull: false,
    autoIncrement: false,
    primaryKey: true,
    type: DataTypes.UUID
  },
  path: {
    type: 'VARCHAR(500)'
  }
}, {
  sequelize,
  freezeTableName: true
});
(async () => {
  await sequelize.sync({ force: true });
  // Code here
})();

module.exports = File

and i am using this model as

const Op = Sequelize.Op;
const file = require('./models/file')
file.findOne({
    where:{
        path:{
            [Op.like]:'%-1001_1101.xml'
        }
    }
}).then((result) => {
    console.log(JSON.stringify(result))
}).catch((e) => {
    console.log(e)
})

This is producing following query

SELECT [uuid], [path], [createdAt], [updatedAt] FROM [File] AS [File] WHERE [File].[path] LIKE N'%-1001_1101.xml' ORDER BY [File].[uuid] OFFSET 0 ROWS FETCH NEXT 1 ROWS ONLY

The issue is with the N char after like in query. How to fix this?

[ephys]

@nareshy-quisitive That part is tracked over here #9107 but we won't be able to do that until we have a proper non-n varchar type, which I opened a RFC for here #14259