A tiny and temporary db for quick data cleansing and transformation. It is a high-level Python coding and would help any Pythonistas up to speed with ETL work.
- Python 3.7 or higher
- rapidfuzz
- openpyxl
pip install bintangfrom bintang import Bintang # import the package
bt = Bintang() # bintang object created
bt.create_table('Person')
print(bt )
# {
# "name": null,
# "tables": [
# "Person"
# ]
# }Use insert function to populate a table. The parameters are record and columns.
# Create a couple of tables and use insert function to insert data.
bt.create_table('Person')
bt['Person'].insert([1,'John','Smith','1 Station St'], ['id','name','surname','address'])
bt['Person'].insert([2,'Jane','Brown','Digging','8 Parade Rd'], ['id','name','surname','hobby','address'])
bt['Person'].insert([3,'Okie','Dokie','7 Ocean Rd'], ['id','name','surname','Address'])
bt['Person'].insert((4,'Maria','Digging','7 Heaven Ave'), ('id','name','hobby','Address'))
bt['Person'].insert((5,'Bing','Digging',None), ('id','name','hobby','Address'))
bt.create_table("FishingClub")
bt['FishingClub'].insert(['Ajes','Freeman','Active'], ['FirstName','LastName','Membership'])
bt['FishingClub'].insert(['John','Smith','Active'], ['FirstName','LastName','Membership'])
bt['FishingClub'].insert(['Jane','Brown','Active'], ['FirstName','LastName','Membership'])
bt['FishingClub'].insert(['Nutmeg','Spaniel','Active'], ['FirstName','LastName','Membership'])
bt['FishingClub'].insert(['Zekey','Pokey','Active'], ['FirstName','LastName','Membership'])Loop your data using iterrows function. This will loop through all the rows one by one in a Python dict.
for idx, row in bt['Person'].iterrows():
print(idx, row)
# 0 {'id': 1, 'name': 'John', 'surname': 'Smith', 'address': '1 Station St', 'hobby': None}
# 1 {'id': 2, 'name': 'Jane', 'surname': 'Brown', 'address': '8 Parade Rd', 'hobby': 'Digging'}
# 2 {'id': 3, 'name': 'Okie', 'surname': 'Dokie', 'address': '7 Ocean Rd', 'hobby': None}
# 3 {'id': 4, 'name': 'Maria', 'surname': None, 'address': '7 Heaven Ave', 'hobby': 'Digging'}If the table is small, you can use print() function to output the records to terminal.
Person.print()
# Table: Person
# --------------+---------+-----------+----------------+-----------
# id | name | surname | address | hobby
# --------------+---------+-----------+----------------+-----------
# 1 | John | Smith | 1 Station St | None
# 2 | Jane | Brown | 8 Parade Rd | Digging
# 3 | Okie | Dokie | 7 Ocean Rd | None
# 4 | Maria | None | 7 Heaven Ave | Digging
# 5 | Bing | None | None | Digging
# --------------+---------+-----------+----------------+-----------
# (5 rows)You should notice that all columns now have been normalised for each row, even though all records have not used all column names during insertion.When the 1st record (idx 0) is inserted, four columns created. id, name, age and address. When insert 4th record is inserted (idx 3), a hobby column added.
Inspect Person table column list. You can also use function table.get_columns() to list all columns.
print(bt['Person'])
# {
# "table name": "Person",
# "columns": [
# {"id": 0,"name": "id"},
# {"id": 1,"name": "name"},
# {"id": 2,"name": "age"},
# etc...
# }Use update function to change the data. The function signature is table.update(column, value, where=None). The value and where parameters can use lambda function argument for smarter expression.
bt['Person'].update('age', 10, where=lambda row: row['name']=='Maria')Retrieve the row stored at the 3rd index by passing 3 inside the list next to table object.
bt['Person'][3]
# {'id': 4, 'name': 'Maria', 'age': 10, 'address': '7 Heaven Ave', 'hobby': 'Digging'}We are going to provide some functions that may be needed most when working with Bintang objects.
Read an Excel file and create Bintang tables. The first row will be assumed as column header. Go to Bintang.Table.read_excel() to read a single sheet and populate the data into created table.
| path: | an excel file path to read from. |
|---|---|
| sheetname: | a list of sheets that only needed to be read. If not specified all available sheets will be read (default). |
bt.read_excel('/path/to/file.xlsx')Read JSON string and create a table or more according to hierarchy paths contained in json 'object'.
| json_str: | a json string |
|---|---|
| tablepaths: | a list of paths which contain a list of objects (equivalent to records). |
# other module import
# ...
import bintang
import json
# example json data
json_str = '{"Page:": 100, "Time": "2033-09-05T00:00:00Z", \
"Person": [{"id": 1, "name": "John", "surname": "Smith", \
"Address": {"number": 1, "street": "Station", "street_type": "Street"}}, \
{"id": 2, "name": "Jane", "surname": "Brown", \
"Address": {"number": 8, "street": "Parade", "street_type": "Road"}}], \
"PersonDetails": [{"person_id": "1", "hobby": "Blogging", "is_meat_eater": true}, \
{"person_id": "2", "hobby": "Reading", "is_meat_eater": null, \
"LuckyDays": [13, 17, 19]}]}'
bt = bintang.Bintang('From JSON')
bt.read_json(json_str)
print(bt) # show bt tables
# {
# "name": "From JSON",
# "tables": [
# "/",
# "/Person",
# "/Person/Address",
# "/PersonDetails",
# "/PersonDetails/LuckyDays"
# ]
# }
# loop through root table ('/')
for idx, row in bt['/'].iterrows():
print(idx, row)
# 0 {'Page:': 100, 'Time': '2033-09-05T00:00:00Z'}
# loop through /Person table.
for idx, row in bt['/Person'].iterrows():
print(idx, row)
# 0 {'Person': 0, 'id': 1, 'name': 'John', 'surname': 'Smith'}
# 1 {'Person': 1, 'id': 2, 'name': 'Jane', 'surname': 'Brown'}
# print /Person/Address table. Because this table under /Person, then each record will have their own
# reference to /Person table.
bt['/Person/Address'].print()
# Table: /Person/Address
# -----------+--------------+--------------+-----------+---------------
# Address | Person | number | street | street_type
# -----------+--------------+--------------+-----------+---------------
# Address | 0 | 1 | Station | Street
# Address | 1 | 8 | Parade | Road
# -----------+--------------+--------------+-----------+---------------
# (2 rows)Please note that since json can contain complex hierarchy paths and still valid (eg. system configuration), then a further transformation is required. A well written JSON can be transformed to Bintang tabular model straight away.
Return one or more columns from lookup table.
| lkp_table: | lookup table |
|---|---|
| on: | lookup key tuples |
| ret_columns: | lookup columns to be returned |
# using tables from Example of Usage section above.
bt['Person'].blookup('FishingClub')], \
[('name','FirstName')], \
['Membership'])
# check results
for idx, row in bt['Person'].iterrows(['name','Membership']):
print(idx, row)
# 0 {'name': 'John', 'Membership': 'Active'}
# 1 {'name': 'Jane', 'Membership': 'Active'}
# 2 {'name': 'Okie', 'Membership': None}
# 3 {'name': 'Maria', 'Membership': None}We can see only John and Jane got the membership because their names exists in both tables.
Bintang.Table.groupby(columns, drop_none=True, group_count=False, counts=None, sums=None, mins=None, maxs=None, means=None, group_concat=None)
Return grouped rows based upon the value of columns.
| columns: | a list of columns that is used to group the data. |
|---|---|
| drop_none: | if True, to drop/exclude the group if all columns hold None. |
| group_count: | if True, create row count from group columns. |
| group_concat: | a column to create group_concat like mysql. |
| counts: | a list of columns to create count column(s) |
| sums: | a list of columns to create sum column(s) |
| mins: | a list of columns to create min column(s) |
| maxs: | a list of columns to create max column(s) |
bt.create_table('Product') # this will be our basis table for grouping
p = bt['Product'] # assign p as the table
p.insert({'id':1, 'brand': 'Shimano', 'class':'rod', 'name':'Extraction','price':299})
p.insert({'id':2, 'brand': 'Shimano', 'class':'rod', 'name':'Zodias Travel','price':399})
p.insert({'id':3, 'brand': 'Ugly Stik', 'class':'rod', 'name':'Balance II','price':63.99})
p.insert({'id':4, 'brand': 'Shimano', 'class':'rod', 'name':'Zodias Travel','price':399})
p.insert({'id':5, 'brand': 'Shimano', 'class':'reel', 'sub class': 'spinning', 'name':'Sedona F1','price':99.00})
p.insert({'id':6, 'brand': 'Shimano', 'class':'reel', 'sub class':'spinning', 'name':'FX Series 4000','price':54.99})
grouped = p.groupby(['brand', 'class'], group_count=True)
grouped.print()
# Table: grouped
# -------------+---------+---------------
# brand | class | group_count
# -------------+---------+---------------
# Shimano | rod | 3
# Ugly Stik | rod | 1
# Shimano | reel | 2
# -------------+---------+---------------
# (3 rows)
grouped = p.groupby(['brand'], group_concat='id', sums=['price']) # another example
grouped.print()
# Table: grouped
# -------------+-------------------+-------------------
# brand | group_concat | sum_price
# -------------+-------------------+-------------------
# Shimano | [1, 2, 4, 5, 6] | 1250.99
# Ugly Stik | [3] | 63.99
# -------------+-------------------+-------------------
# (2 rows)Return a new table from an inner join operation.
| right_table: | name of right table or the second table. |
|---|---|
| on: | a list of pair columns used for the join. |
| into: | a new table name to hold the result. |
| out_leftcolumns: | columns output from left table. If not specified then it will return all columns. |
| out_rightcolumns: | columns outpout from right table. If not specified then it will return all columns. |
bt.create_table('Person') # This will be a left table
# insert some record here. See insert below for an example.
# ...
bt.create_table('FishingClub') # this will be a right table
# insert some records here. See insert below for an example.
# ...
# let's match the two tables for their firt name and last name.
res = bt.innerjoin('Person' # left table
,'FishingClub' # right table
,[('name','FirstName'), ('surname','LastName')] # on
,into='Fisherman'
,out_lcolumns=['name','address']
,out_rcolumns=['Membership']
)
# check the result. you can loop through 'Fisherman' or res.
for idx, row in bt['Fisherman'].iterrows():
print(idx, row)Insert a record into a table.
| record: | a list/tuple of data. Or a dict where key=column, value=record |
|---|---|
| columns: | a list/tuple of columns (in the same order as in the record) |
bt.create_table('Person')
p = bt.get_table('Person') # get table object for Person
# insert data directly from table object instead throug bt object.
p.insert([1,'John','Smith','1 Station St'], ['id','name','surname','address'])
p.insert([2,'Jane','Brown','Digging','8 Parade Rd'], ['id','name','surname','hobby','address'])
p.insert([3,'Okie','Dokie','7 Ocean Rd'], ['id','name','surname','Address'])
p.insert((4,'Maria','Digging','7 Heaven Ave'), ('id','name','hobby','Address'))
p.insert((5,'Bing','Digging',None), ('id','name','hobby','Address'))
bt.create_table('FishingClub')
# lets make a list of columns so we can pass it to insert.
columns = ['FirstName','LastName','Membership']
bt['FishingClub'].insert(['Ajes','Freeman','Active'], columns)
bt['FishingClub'].insert(['John','Smith','Active'], columns)
bt['FishingClub'].insert(['John','Brown','Active'], columns)
bt['FishingClub'].insert(['Okie','Dokie','Active'], columns)
bt['FishingClub'].insert(['Zekey','Pokey','Active'], columns)
bt.create_table("Product")
prod = bt['Product']
# example of assigning a dictionary argument for record parameter.
prod.insert({'id':1, 'name':'Hook','price':1.60})
prod.insert({'id':2, 'name':'Sinker','price':1.20})
prod.insert({'id':3, 'name':'Reels','price':75})Loop through Bintang table's rows and yield index and row. Row can be called out as dict (default) or list.
| columns: | a list of columns to output. If None, will output all columns. |
|---|---|
| row_type: | either 'dict' (default) or 'list'. |
for idx, row in bt['tablename'].iterrows():
# do something with idx or row
print(idx, row)Read csv file and populate its records to table.
| path: | a csv file path to read from. |
|---|---|
| delimiter: | field seperator, by default it'll accept a comma character. |
| header_row: | the row number that contains column name or label. |
## example data in csv file
# "id","name","surname","address","hobby"
# "1","John","Smith","1 Station St",""
# "2","Jane","Brown","8 Parade Rd","Digging"
bt.create_table('Person')
bt['Person'].read('/path/to/file.csv')
bt['Person'].print()
# Table: Person
# ------+---------+-----------+----------------+-----------
# id | name | surname | address | hobby
# ------+---------+-----------+----------------+-----------
# 1 | John | Smith | 1 Station St |
# 2 | Jane | Brown | 8 Parade Rd | Digging
# ------+---------+-----------+----------------+-----------
# (2 rows)Read an Excel file into Bintang table.
| path: | an excel file path to read from. |
|---|---|
| sheetname: | the sheetname to read from. |
| header_row: | the row number that contains column name or label. |
bt.create_table('Person')
bt['Person'].read_excel('/path/to/file.xlsx', 'Sheet1')Read sql table and populate the data to Bintang table.
| conn: | pyodbc database connection |
|---|---|
| sql_str: | sql query, if none it will select * from a same sql table name. |
| params: | sql parameters |
import bintang
import pyodbc
# connect to sql server
conn_str = "DRIVER={ODBC Driver 17 for SQL Server};SERVER=localhost;PORT=1443;DATABASE=test;Trusted_Connection=yes;"
conn = pyodbc.connect(conn_str)
sql_str = "SELECT * FROM Person WHERE LastName=?"
params = ('Dokey')
bt = bintang.Bintang()
bt.create_table('Person')
bt['Person'].read_sql(conn, sql_str, params)
for idx, row in bt['Person'].iterrows():
print(idx, row)
# would print {'ID': 3, 'FirstName': 'Okie', 'LastName': 'Dokey', 'address': '7 Ocean Rd'}
conn.close()Write bintang table to a csv file.
| path: | a csv file path to write to. |
|---|---|
| index: | write row index if it sets True. |
| delimiter: | field seperator |
| quotechar: | a character to quote the data |
| quoting: | the csv enum for quoting, csv.QUOTE_MINIMAL or 0, csv.QUOTE_ALL or 1, csv.QUOTE_NONNUMERIC or 2, csv.QUOTE_NONE or 3 |
Notes: setting quoting parameter properly will provide correct value to be presented in csv even if the value containing a delimiter character.
bt['tablename'].to_csv('/path/to/file.csv')Write Bintang table to an Excel file.
| path: | an excel file path to write to. |
|---|---|
| index: | write row index if it sets True. |
bt['tablename'].to_excel('/path/to/file.xlsx')This is just a placeholder. Python make it easy when serializing a dict object to JSON. Conversion would be done by built-in json.JSONEncoder(). Here an example of using our to_dict() function then use build-in module json to convert/export dict to JSON.
# other modules here
# ...
import json
# other codes here
# ...
dict_obj = bt['table_name'].to_dict()
# example to serialise dict_obj to json string
json_str = json.dumps(dict_obj)
# use json_str here!
# ...
# example to write dict_obj to a json file
with open ('myfile.json', 'w') as fp:
json.dump(dict_obj, fp) # this would serialise dict_obj into myfile.jsonInsert records into sql table. Notes: Currently tested for SQL Server 15 and Postgresql 16. However this function should work with other dbms supported by pyodbc.
| conn: | pyodbc database connection |
|---|---|
| table: | the table name in the sql database |
| columns: | If a dictionary then a columns mapping where the key is sql column (destination) and the value is bintang columns (source). If a list, column mapping will be created automatically assuming source columns and destination columns are the same. If not provided it assumes that user wants to insert all the columns from the table. |
| schema: | the schema name the sql table belongs to. |
| method: | 'prep' to use prepared statement (default) or 'string' to use sql string. To avoid sql injection, never use string method when the datasource is external or not known. |
| max_rows: | maximum rows per batch insert. Allowed value would be from 1 to 1000. Insert more then 1 record when using prep require all data in a column to use the same type, otherwise will raise error. |
import bintang
import pyodbc
bt = bintang.Bintang('my bintang')
bt.create_table('Person')
person = bt.get_table('Person')
person.insert([1,'John','Smith','1 Station St'], ['id','name','surname','address'])
person.insert([2,'Jane','Brown','Digging','8 Parade Rd'], ['id','name','surname','address'])
person.insert([3,'Okie','Dokey','7 Ocean Rd'], ['id','name','surname','address'])
person.insert((4,'Maria','Digging','7 Heaven Ave'), ('id','name','hobby','Address'))
person.insert((5,'Bing','Digging',None), ('id','name','hobby','Address'))
# let's map column ID, FirstName, LastName, Address in database to bintang's Person table.
columns = {'ID':'id', 'FirstName':'name', 'LastName':'surname', 'Address':'address'}
# connect to database
conn = pyodbc.connect("DRIVER={ODBC Driver 17 for SQL Server};SERVER=localhost;PORT=1443;DATABASE=test;Trusted_Connection=yes;")
# send data to sql
ret = person.to_sql(conn, 'Person', columns)
print(f'{ret} record(s) affected.')
conn.commit()
conn.close()Add a table to Bintang object. Think Bintang is a container of tables.
| table_object: | table object is a table created from create_table function. |
|---|
Create a table inside Bintang object
| name: | string to name the table |
|---|---|
| columns: | a list of columns. Bintang can create columns during record insertion (dynamic schema effect). |
Drop table from tables container.
| name: | table name |
|---|
bt.drop_table(name)To rename Bintang object name can be done through normal assignment.
import bintang
bt = bintang.Bintang('my db')
bt.name = 'your db' # change name from 'my db' to 'your db'Add a column to table. Bintang can create columns during record insertion (dynamic schema effect).
| column: | a string of column name |
|---|
To delete row(s) that meets where conditions.
| where: | a lambda expression |
|---|
# assume p is a Person table
p.delete(lambda row: row['name'].lower()=='john')Notes: lambda is case-sensitive.
To delete the row at index.
| index: | row index |
|---|
Drop a column from a table.
| column: | a string of column name |
|---|
Return a list of columns.
Return a row at given index.
| index: | row index. |
|---|---|
| columns: | a list of specific columns wanted in the row. |
| row_type: | row can be in a 'list' or 'dict' (default) |
Return True if index exists otherwise None.
| index: | row index. |
|---|
To rename table name, can be done through normal assignment.
bt['my table'].name = 'your table' # change name from 'my table' to 'your table'Print rows to terminal in table format. This would be handy if the table can fit into terminal.
| columns: | a list of columns to output. If None, will output all columns (default). |
|---|---|
| show_data_type: | if True, will output data type. |
# assume Person table object exists and has data
Person.print()Rename column name.
| old_column: | the existing column that will be renamed. |
|---|---|
| new_column: | new column name. |
To update row(s) in the table
| column: | column to be updated. If column does not exist then it will be created (dynamic schema effect). |
|---|---|
| value: | the value to be set for the row(s). Can also use lambda. |
| where: | condition in lambda so which row(s) that needs update. |
# assume Person table object exists and has records.
Person.update('job', 'fisherman') # this will create job column and set value of the rows to 'fisherman'
# update only row that has name 'John'
Person.update('job', 'fisherman', where=lambda row: row['name']=='John')
# create 'full name' column and populate value with name and surname combined.
bt['Person'].update('full name', lambda row: row['name'] + ' ' + row['surname'])
# one liner conditional expression is common with lambda.
# for instance, if you know there is None in surname then you want to resolve the full name as name.
bt['Person'].update('full name', lambda row: row['name'] if row['surname'] is None else row['name'] + ' ' + row['surname'])To update column's attributes, eg. data_type, column_size and ordinal_position.
| name: | the column name |
|---|---|
| data_type: | python data type eg. str, int, float |
| column_size: | string length if data_type is a str |
| ordinal_position: | integer for column order. |
Note: At the moment data_type and column_size are used for generating sql table and not used for presenting the data.
To update the row at idx. So only one row will be affected.
| index: | row index |
|---|---|
| column: | which column will be updated |
| value: | new value |