dbtoyaml - Database to YAML


dbtoyaml – extract the schema of a PostgreSQL database in YAML format


dbtoyaml [option...] dbname


dbtoyaml is a utility for extracting the schema of a PostgreSQL database to a YAML formatted specification. By default, the specification is output as a single output stream, which can be redirected or explicitly sent to a file. As an alternative, the --multiple-files option allows you to break down the specification into multiple files, one for each object (see Multiple File Output).

Note that JSON is an official subset of YAML version 1.2, so the dbtoyaml output should also be compatible with JSON tools.

The output format is as follows:

schema public:
  owner: postgres
  - postgres:
    - all
    - all
  table t1:
      check_expr: (c2 > 123)
      - c2
    - c1:
        not_null: true
        type: integer
    - c2:
        type: smallint
    - c3:
        default: 'false'
        type: boolean
    - c4:
        type: text
        - c2
          - c21
          schema: s1
          table: t2
    owner: alice
        - c1
schema s1:
  owner: bob
  - bob:
    - all
  - alice:
    - all
  table t2:
    - c21:
         not_null: true
         type: integer
    - c22:
         type: character varying(16)
    owner: bob
        - c21
    - bob:
      - all
    - PUBLIC:
      - select
    - alice:
      - insert:
          grantable: true
      - delete:
          grantable: true
      - update:
          grantable: true
    - carol:
        grantor: alice
        - insert

The above should be mostly self-explanatory. The example database has two tables, named t1 and t2, the first –owned by user ‘alice’– in the public schema and the second –owned by user ‘bob’– in a schema named s1 (also owned by ‘bob’). The columns: specifications directly under each table list each column in that table, in the same order as shown by PostgreSQL. The specifications primary_key:, foreign_keys: and check_constraints: define PRIMARY KEY, FOREIGN KEY and CHECK constraints for a given table. Additional specifications (not shown) define unique constraints and indexes.

User ‘bob’ has granted all privileges to ‘alice’ on the s1 schema. On table t2, he also granted SELECT to PUBLIC; INSERT, UPDATE and DELETE to ‘alice’ with GRANT OPTION; and she has in turn granted INSERT to user ‘carol’.

dbtoyaml currently supports extracting information about nearly all types of PostgreSQL database objects. See API Reference for a list of supported objects.

Multiple File Output

The --multiple-files option breaks down the output into multiple files under a given root directory. The root is created if it does not exist. The root directory name defaults to metadata in the system configuration file. The location of the root directory defaults to the configuration item repository.path or can be specified using the --repository option (see Configuration and Common Command Line Options for further details).

The first level contains schema.<name> subdirectories, schema.<name>.yaml files and <objtype>.<name>.yaml files, where <name> is the name of the corresponding objects and <objtype> is the type of top-level (non-schema) object. Note that non-schema refers to PostgreSQL extensions, casts, languages or foreign data wrappers.

The second level, i.e., the schema.<name> subdirectories contain <objtype>.<name>.yaml files for each object in the particular schema (but see below for caveats).

Object Name Conflicts

The names of PostgreSQL objects can include characters that are not allowed in filesystem object names. The most common example is the division operator (‘/’), but even table names can include non-alphanumeric characters, if the identifiers are quoted.

In addition, one can define two or more objects with the same base name, e.g., function foo(integer) and function foo(text), or a table named "My Table" and another named "my table" or "MY TABLE". On certain operating systems, i.e., Windows, it is not possible to create two files in the same directory that differ only in the case of their characters.

In order to deal with the aforementioned issues, dbtoyaml places certain objects in common files and transforms object identifiers so that they are suitable for use in files and directories. For example, the information for all user-defined casts are written to the file cast.yaml in the root directory. Functions with the same name but different arguments are written to a single file, e.g., function.foo.yaml in the first example above. Identifiers are also converted to all lowercase, non-alphanumeric characters (excluding underscore) are converted to underscores and, by default, schema object names are truncated to 32 characters.

If two object names, thus transformed, map to the same string, then the objects’ information is written to the same file, e.g., table.my_table.yaml in the second example above. If you prefer to change the default truncation length, please define the environment variable PYRSEAS_MAX_IDENT_LEN to some integer value (up to 63).

Version Control and Dropped Objects

It is expected that the output of dbtoyaml --multiple-files will be placed under version control. Further invocations should then update the files in the same directory tree. However, if an object is dropped from the database dbtoyaml would normally only output files for new or changed objects–and thus keep the dropped object file under version control. To deal with dropped objects, dbtoyaml -d outputs a special YAML “index” file, named database.<dbname>.yaml in the root directory. When dbtoyaml -d is run a second time, it looks for this “index” file and if found, proceeds to delete the previous run’s .yaml files before outputting new ones.


dbtoyaml accepts the following command-line arguments (in addition to the Common Command Line Options):


Specifies the name of the database whose schema is to extracted.
-m, --multiple-files

Extracts the schema to a two-level directory tree. See Multiple File Output above.

-n <schema>
--schema <schema>

Extracts only a schema matching schema. By default, all schemas are extracted. Multiple schemas can be extracted by using multiple -n switches. Note that normally all objects that belong to the schema are extracted as well, unless excluded otherwise.

-N <schema>
--exclude-schema <schema>

Does not extract schema matching schema. This can be given more than once to exclude several schemas.

-O, --no-owner

Do not output object ownership information. By default, as seen in the sample output above, database objects (schemas, tables, etc.) that can be owned by some user, are shown with an “owner: username” element. The -O switch suppresses all those lines.

-t <table>
--table <table>

Extract only tables matching table. Multiple tables can be extracted by using multiple -t switches. Note that selecting a table may cause other objects, such as an owned sequence, to be extracted as well

-T <table>
--exclude-table <table>

Do not extract tables matching table. Multiple tables can be excluded by using multiple -T switches.

-x, --no-privileges

Do not output access privilege information. By default, as seen in the sample output above, if specific GRANTs have been issued on various objects (schemas, tables, etc.), the privileges are shown under each object. The -x switch suppresses all those lines.


To extract a database called moviesdb into a file:

dbtoyaml moviesdb > moviesdb.yaml

To extract only the schema named store:

dbtoyaml --schema=store moviesdb > moviesdb.yaml

To extract the tables named film and genre:

dbtoyaml -t film -t genre moviesdb -o moviesdb.yaml

To extract objects, to standard output, except those in schemas product and store:

dbtoyaml -N product -N store moviesdb

To extract objects to a directory under version control:

dbtoyaml moviesdb -d movies/dbspec