loadData

Loads data from a CSV file into an existing table when you add it to your changelog.

Uses

A value of NULL in a cell will be converted to a database NULL rather than the string ‘NULL'. Lines starting with a number sign (#) are treated as comments. You can change the comment pattern by specifying commentLineStartsWith attribute. To disable comments, set commentLineStartsWith to empty value.

If the data type for a load column is set to NUMERIC, numbers are parsed in the US locale (for example: 123.45). Date/Time values included in the CSV file should be in ISO format to be parsed correctly by Liquibase. Liquibase initially sets the date format to ‘yyyy-MM-dd'T'HH:mm:ss' and then it checks for two special cases which will override the data format string:

  • If the string representing the date/time includes a period (.), then the date format is changed to ‘yyyy-MM-dd'T'HH:mm:ss.SSS'.
  • If the string representing the date/time includes a space, then the date format is changed to ‘yyyy-MM-dd HH:mm:ss'.

Once the date format string is set, Liquibase will then call the SimpleDateFormat.parse() method attempting to parse the input string so that it can return a date/time. If problems occur, then a ParseException is thrown and the input string is treated as a String for the INSERT command to be generated. If UUID type is used, UUID value is stored as string and NULL in cell is supported.

Loading data with the loadData tag

All CSV columns are used by default while generating SQL even if they are not described in columns property. If you want to skip the specific headers in the CSV file, use type property with a skip value.

Imagine that you have a table where a,b,c are column names, and 1,2,3 are values. To load only a and b columns, add a column configuration for it and set its type to "skip":

<column name="c" header="c" type="skip" />

If you use the generateChangeLog command to re-create the current state of the database and receive the output of this data as the CSV file in a specific directory, add dataOutputDirectory where the CSV file with the insert statements will be kept:

liquibase --diffTypes=tables,functions,views,columns,indexes,foreignkeys,primarykeys,uniqueconstraints,data,storedprocedure,triggers,sequences --dataOutputDirectory=data generateChangeLog

If you don't use the --dataOutputDirectory flag while running the command or the loadData tag added in the changelog, you will have the insert statements in your changelog:

 liquibase --diffTypes=tables,functions,views,columns,indexes,foreignkeys,primarykeys,uniqueconstraints,data,storedprocedure,triggers,sequences --changeLogFile=myChangelog.xml generateChangeLog

Available attributes

Name Description Required for Supports Since
catalogName Name of the catalog. all 3.0
commentLineStartsWith Lines starting with this are treated as comment and ignored. all
encoding Encoding of the CSV file (defaults to UTF-8). all
file CSV file to load. all all
quotchar The quote character for string fields containing the separator character. all
relativeToChangelogFile Whether the file path relative to the root changelog file rather than to the classpath. all
schemaName Name of the schema. all
separator Character separating the fields. all
tableName Name of the table to insert data into. all all
usePreparedStatements Use prepared statements instead of insert statement strings if the database supports it. all

Nested properties

Name Description Required for Supports Multiple allowed
columns / column Column mapping and defaults can be defined. The header or index attributes need to be defined. If the header name in the CSV is different than the column, name needs to be inserted. If no column is defined at all, the type is taken from a database. Otherwise, for non-string columns the type definition might be required. all yes

Nested property attributes

Name Name of the Column (Required)
type Data type of the column. Its value has to be one of the LOAD_DATA_TYPE. If you want to skip loading a specific column, use the skip data type that was mentioned earlier, otherwise all columns in the CSV file will be used.
header Name of the column in the CSV file from which the value for the column will be taken if its different from the column name. Ignored if index is also defined.
index Index of the column in the CSV file from which the value for the column will be taken.

Database support

Database Notes Auto rollback
DB2/LUW Supported No
DB2/z Supported No
Derby Supported No
Firebird Supported No
H2 Supported No
HyperSQL Supported No
INGRES Supported No
Informix Supported No
MariaDB Supported No
MySQL Supported No
Oracle Supported No
PostgreSQL Supported No
SQL Server Supported No
SQLite Supported No
Sybase Supported No
Sybase Anywhere Supported No