Simple DDL Parser
Simple DDL Parser to parse SQL (HQL, TSQL, AWS Redshift, Snowflake and other dialects) ddl files to json/python dict with full information about columns: types, defaults, primary keys, etc.
Build with ply (lex & yacc in python). A lot of samples in 'tests/.
Is it Stable?
Yes, library already has about 5000+ usage per day - https://pypistats.org/packages/simple-ddl-parser.
As maintainer I guarantee that any backward incompatible changes will not be done in patch or minor version. Only additionals & new features.
However, in process of adding support for new statements & features I see that output can be structured more optimal way and I hope to release version 1.0.* with more struct output result. But, it will not be soon, first of all, I want to add support for so much statements as I can. So I don't think make sense to expect version 1.0.* before, for example, version 0.26.0 :)
How does it work?
Parser tested on different DDLs for PostgreSQL & Hive. But idea to support as much as possible DDL dialects (AWS Redshift, Oracle, Hive, MsSQL, etc.). You can check dialects sections after Supported Statements section to get more information that statements from dialects already supported by parser.
If you need some statement, that not supported by parser yet: please provide DDL example & information about that is it SQL dialect or DB.
Types that are used in your DB does not matter, so parser must also work successfuly to any DDL for SQL DB. Parser is NOT case sensitive, it did not expect that all queries will be in upper case or lower case. So you can write statements like this:
Alter Table Persons ADD CONSTRAINT CHK_PersonAge CHECK (Age>=18 AND City='Sandnes');
It will be parsed as is without errors.
If you have samples that cause an error - please open the issue (but don't forget to add ddl example), I will be glad to fix it.
A lot of statements and output result you can find in tests on the github - https://github.com/xnuinside/simple-ddl-parser/tree/main/tests .
How to install
pip install simple-ddl-parser
How to use
Extract additional information from HQL (& other dialects)
In some dialects like HQL there is a lot of additional information about table like, fore example, is it external table, STORED AS, location & etc. This propertie will be always empty in 'classic' SQL DB like PostgreSQL or MySQL and this is the reason, why by default this information are 'hidden'.
Also some fields hidden in HQL, because they are simple not exists in HIVE, for example 'deferrable_initially'
To get this 'hql' specific details about table in output please use 'output_mode' argument in run() method.
example:
ddl = """
CREATE TABLE IF NOT EXISTS default.salesorderdetail(
SalesOrderID int,
ProductID int,
OrderQty int,
LineTotal decimal
)
PARTITIONED BY (batch_id int, batch_id2 string, batch_32 some_type)
LOCATION 's3://datalake/table_name/v1'
ROW FORMAT DELIMITED
FIELDS TERMINATED BY ','
COLLECTION ITEMS TERMINATED BY '\002'
MAP KEYS TERMINATED BY '\003'
STORED AS TEXTFILE
"""
result = DDLParser(ddl).run(output_mode="hql")
print(result)
And you will get output with additional keys 'stored_as', 'location', 'external', etc.
# additional keys examples
{
...,
'location': "'s3://datalake/table_name/v1'",
'map_keys_terminated_by': "'\\003'",
'partitioned_by': [{'name': 'batch_id', 'size': None, 'type': 'int'},
{'name': 'batch_id2', 'size': None, 'type': 'string'},
{'name': 'batch_32', 'size': None, 'type': 'some_type'}],
'primary_key': [],
'row_format': 'DELIMITED',
'schema': 'default',
'stored_as': 'TEXTFILE',
...
}
If you run parser with command line add flag '-o=hql' or '--output-mode=hql' to get the same result.
Possible output_modes: ["mssql", "mysql", "oracle", "hql", "sql", "redshift", "snowflake"]
From python code
from simple_ddl_parser import DDLParser
parse_results = DDLParser("""create table dev.data_sync_history(
data_sync_id bigint not null,
sync_count bigint not null,
sync_mark timestamp not null,
sync_start timestamp not null,
sync_end timestamp not null,
message varchar(2000) null,
primary key (data_sync_id, sync_start)
); """).run()
print(parse_results)
To parse from file
from simple_ddl_parser import parse_from_file
result = parse_from_file('tests/sql/test_one_statement.sql')
print(result)
From command line
simple-ddl-parser is installed to environment as command sdp
sdp path_to_ddl_file
# for example:
sdp tests/sql/test_two_tables.sql
You will see the output in schemas folder in file with name test_two_tables_schema.json
If you want to have also output in console - use -v flag for verbose.
sdp tests/sql/test_two_tables.sql -v
If you don't want to dump schema in file and just print result to the console, use --no-dump flag:
sdp tests/sql/test_two_tables.sql --no-dump
You can provide target path where you want to dump result with argument -t, --targer:
sdp tests/sql/test_two_tables.sql -t dump_results/
More details
DDLParser(ddl).run()
.run() method contains several arguments, that impact changing output result. As you can saw upper exists argument output_mode that allow you to set dialect and get more fields in output relative to chosen dialect, for example 'hql'. Possible output_modes: ["mssql", "mysql", "oracle", "hql", "sql"]
Also in .run() method exists argument group_by_type (by default: False). By default output of parser looks like a List with Dicts where each dict == one entitiy from ddl (table, sequence, type, etc). And to understand that is current entity you need to check Dict like: if 'table_name' in dict - this is a table, if 'type_name' - this is a type & etc.
To make work little bit easy you can set group_by_type=True and you will get output already sorted by types, like:
{
'tables': [all_pasrsed_tables],
'sequences': [all_pasrsed_sequences],
'types': [all_pasrsed_types],
'domains': [all_pasrsed_domains],
...
}
For example:
ddl = """
CREATE TYPE "schema--notification"."ContentType" AS
ENUM ('TEXT','MARKDOWN','HTML');
CREATE TABLE "schema--notification"."notification" (
content_type "schema--notification"."ContentType"
);
CREATE SEQUENCE dev.incremental_ids
INCREMENT 10
START 0
MINVALUE 0
MAXVALUE 9223372036854775807
CACHE 1;
"""
result = DDLParser(ddl).run(group_by_type=True)
# result will be:
{'sequences': [{'cache': 1,
'increment': 10,
'maxvalue': 9223372036854775807,
'minvalue': 0,
'schema': 'dev',
'sequence_name': 'incremental_ids',
'start': 0}],
'tables': [{'alter': {},
'checks': [],
'columns': [{'check': None,
'default': None,
'name': 'content_type',
'nullable': True,
'references': None,
'size': None,
'type': '"schema--notification"."ContentType"',
'unique': False}],
'index': [],
'partitioned_by': [],
'primary_key': [],
'schema': '"schema--notification"',
'table_name': '"notification"'}],
'types': [{'base_type': 'ENUM',
'properties': {'values': ["'TEXT'", "'MARKDOWN'", "'HTML'"]},
'schema': '"schema--notification"',
'type_name': '"ContentType"'}]}
ALTER statements
Right now added support only for ALTER statements with FOREIGEIN key
For example, if in your ddl after table defenitions (create table statements) you have ALTER table statements like this:
ALTER TABLE "material_attachments" ADD FOREIGN KEY ("material_id", "material_title") REFERENCES "materials" ("id", "title");
This statements will be parsed and information about them putted inside 'alter' key in table's dict.
For example, please check alter statement tests - tests/test_alter_statements.py
More examples & tests
You can find in tests/ folder.
Dump result in json
To dump result in json use argument .run(dump=True)
You also can provide a path where you want to have a dumps with schema with argument .run(dump_path='folder_that_use_for_dumps/')
Supported Statements
CREATE TABLE [ IF NOT EXISTS ] + columns defenition, columns attributes: column name + type + type size(for example, varchar(255)), UNIQUE, PRIMARY KEY, DEFAULT, CHECK, NULL/NOT NULL, REFERENCES, ON DELETE, ON UPDATE, NOT DEFERRABLE, DEFERRABLE INITIALLY, GENERATED ALWAYS, STORED, COLLATE
STATEMENTS: PRIMARY KEY, CHECK, FOREIGN KEY in table defenitions (in create table();)
ALTER TABLE STATEMENTS: ADD CHECK (with CONSTRAINT), ADD FOREIGN KEY (with CONSTRAINT), ADD UNIQUE, ADD DEFAULT FOR
PARTITION BY statement
CREATE SEQUENCE with words: INCREMENT, START, MINVALUE, MAXVALUE, CACHE
CREATE TYPE statement: AS ENUM, AS OBJECT, INTERNALLENGTH, INPUT, OUTPUT
LIKE statement (in this and only in this case to output will be added 'like' keyword with information about table from that we did like - 'like': {'schema': None, 'table_name': 'Old_Users'}).
TABLESPACE statement
COMMENT ON statement
CREATE SCHEMA [IF NOT EXISTS] ... [AUTHORIZATION] ...
CREATE DOMAIN [AS]
CREATE [SMALLFILE | BIGFILE] [TEMPORARY] TABLESPACE statement
CREATE DATABASE + Properties parsing
HQL Dialect statements
- PARTITIONED BY statement
- ROW FORMAT, ROW FORMAT SERDE
- WITH SERDEPROPERTIES ("input.regex" = "..some regex..")
- STORED AS
- COMMENT
- LOCATION
- FIELDS TERMINATED BY, LINES TERMINATED BY, COLLECTION ITEMS TERMINATED BY, MAP KEYS TERMINATED BY
- TBLPROPERTIES ('parquet.compression'='SNAPPY' & etc.)
MSSQL / MySQL/ Oracle
- type IDENTITY statement
- FOREIGN KEY REFERENCES statement
- 'max' specifier in column size
- CONSTRAINT ... UNIQUE, CONSTRAINT ... CHECK, CONSTRAINT ... FOREIGN KEY, CONSTRAINT ... PRIMARY KEY
- CREATE CLUSTERED INDEX
Oracle
- ENCRYPT column property [+ NO SALT, SALT, USING]
- STORAGE column property
AWS Redshift Dialect statements
ENCODE column property
SORTKEY, DISTSTYLE, DISTKEY, ENCODE table properties
CREATE TEMP / TEMPORARY TABLE
syntax like with LIKE statement:
create temp table tempevent(like event);
Snowflake Dialect statements
- CREATE .. CLONE statements for table, database and schema
- CREATE TABLE .. CLUSTER BY ..
- CONSTRAINT .. [NOT] ENFORCED
TODO in next Releases (if you don't see feature that you need - open the issue)
-1. Add base support for BigQuery DDL dialect.
0. Add support for ALTER TABLE ... ADD COLUMN
- Add more support for CREATE type IS TABLE (example: CREATE OR REPLACE TYPE budget_tbl_typ IS TABLE OF NUMBER(8,2);
- Add support (ignore correctly) ALTER TABLE ... DROP CONSTRAINT ..., ALTER TABLE ... DROP INDEX ...
- Add support for COMMENT ON statement
- Add support for SKEWED BY for HQL
non-feature todo
- Provide API to get result as Python Object
- Add online demo (UI) to parse ddl
Historical context
This library is an extracted parser code from https://github.com/xnuinside/fakeme (Library for fake relation data generation, that I used in several work projects, but did not have time to make from it normal open source library)
For one of the work projects I needed to convert SQL ddl to Python ORM models in auto way and I tried to use https://github.com/andialbrecht/sqlparse but it works not well enough with ddl for my case (for example, if in ddl used lower case - nothing works, primary keys inside ddl are mapped as column name not reserved word and etc.).
So I remembered about Parser in Fakeme and just extracted it & improved.
xnuinside
