liquibase : generate changelogs from existing database - liquibase

Question

Is it possible with liquibase to generate changelogs from an existing database?
I would like to generate one xml changelog per table (not every create table statements in one single changelog).

Answer 1

If you look into documentation it looks like it generates only one changelog with many changesets (one for each table). So by default there is no option to generate changelogs per table.

Answer 2

While liquibase generate-changelog still doesn't support splitting up the generated changelog, you can split it yourself.
If you're using JSON changelogs, you can do this with jq.
I created a jq filter to group the related changesets, and combined it with a Bash script to split out the contents. See this blog post
jq filter, split_liquibase_changelog.jq:
# Define a function for mapping a changes onto its destination file name
# createTable and createIndex use the tableName field
# addForeignKeyConstraint uses baseTableName
# Default to using the name of the change, e.g. createSequence
def get_change_group: map(.tableName // .baseTableName)[0] // keys[0];
# Select the main changelog object
.databaseChangeLog
# Collect the changes from each changeSet into an array
| map(.changeSet.changes | .[])
# Group changes according to the grouping function
| group_by(get_change_group)
# Select the grouped objects from the array
| .[]
# Get the group name from each group
| (.[0] | get_change_group) as $group_name
# Select both the group name...
| $group_name,
# and the group, wrapped in a changeSet that uses the group name in the ID and
# the current user as the author
{ databaseChangelog: {
changeSet: {
id: ("table_" + $group_name),
author: env.USER,
changes: . } } }
Bash:
#!/usr/bin/env bash
# Example: ./split_liquibase_changelog.sh schema < changelog.json
set -e -o noclobber
OUTPUT_DIRECTORY="${1:-schema}"
OUTPUT_FILE="${2:-schema.json}"
# Create the output directory
mkdir --parents "$OUTPUT_DIRECTORY"
# --raw-output: don't quote the strings for the group names
# --compact-output: output one JSON object per line
jq \
--raw-output \
--compact-output \
--from-file split_liquibase_changelog.jq \
| while read -r group; do # Read the group name line
# Read the JSON object line
read -r json
# Process with jq again to pretty-print the object, then redirect it to the
# new file
(jq '.' <<< "$json") \
> "$OUTPUT_DIRECTORY"/"$group".json
done
# List all the files in the input directory
# Run jq with --raw-input, so input is parsed as strings
# Create a changelog that includes everything in the input path
# Save the output to the desired output file
(jq \
--raw-input \
'{ databaseChangeLog: [
{ includeAll:
{ path: . }
}
] }' \
<<< "$OUTPUT_DIRECTORY"/) \
> "$OUTPUT_FILE"
If you need to use XML changesets, you can try adapting this solution using an XML tool like XQuery instead.