使用 node.js 从 SQL 服务器读取生僻字

Question

I encounter a problem when I tried to read the data from a ERP database. There is a data called '鉫承工程' and it save in SQL Server like ' 承工程'(data type is varchar). The ERP System can show the correct word as '鉫承工程'.

However, I use node.js to read the data and it shows me the garbled text like '�r承工程'. How can I get the correct word without changing the data in database?

Answer 1

Have a read through the Constants (Transact-SQL) documentation with respect to Character string constants:

Character string constants are enclosed in single quotation marks and include alphanumeric characters (az, AZ, and 0-9) and special characters, such as exclamation point (,), at sign (@). and number sign (#). Character string constants are assigned the default collation of the current database, If the COLLATE clause is used. the conversion to the database default code page still happens before the conversion to the collation specified by the COLLATE clause. Character strings typed by users are evaluated through the code page of the computer and are translated to the database default code page if it is required.

The key statement here is:

Character string constants are assigned the default collation of the current database.

The implication here is that, when a column's collation is different than the database's default collation, using regular character string constants to insert values containing international characters can cause loss of information (ie: corrupted characters) as the literal is first interpreted using the database's default collation and is then converted to the column's collation before storage.

When this happens character corruption is usually evident with unconvertible characters getting replaced by the question mark ( ? ) character.

Consider the following examples:

use master;
-- Typical default collation on systems using en-US...
create database Z_Demo1 collate SQL_Latin1_General_CP1_CI_AS;
create database Z_Demo2 collate Chinese_PRC_CI_AS;
go
use Z_Demo1;
go
create table dbo.Demo (
  ID int not null identity(1,1),
  [鉫承工程] varchar(50) collate Chinese_PRC_CI_AS
);
insert dbo.Demo ([鉫承工程]) values ('承工程');
insert dbo.Demo ([鉫承工程]) values (N'承工程');
select * from dbo.Demo;
go
use Z_Demo2;
go
create table dbo.Demo (
  ID int not null identity(1,1),
  [鉫承工程] varchar(50) collate Chinese_PRC_CI_AS
);
insert dbo.Demo ([鉫承工程]) values ('承工程');
insert dbo.Demo ([鉫承工程]) values (N'承工程');
select * from dbo.Demo;
go
use master;
go
drop database if exists Z_Demo1;
drop database if exists Z_Demo2;
go

Which outputs the following result sets:

ID	鉫承工程
1	???
2	承工程

ID	鉫承工程
1	承工程
2	承工程

The first demo failed to insert both values correctly because the column's collation, Chinese_PRC_CI_AS , was different than the database's default collation, SQL_Latin1_General_CP1_CI_AS .

You can avoid this issue of differing collations by using National Language string constants which are prefixed with an uppercase N character, or from application code specify the column type as NVARCHAR(...length...) instead of VARCHAR(...length...) . Using the above tables from NodeJS using the mssql module, for example, you would specify the data types as sql.NVarChar(50) , eg: request.input("鉫承工程", sql.NVarChar(50), "承工程") .