如何从 KDB 中获取表或视图元数据并保存到数据结构中？

Question

I have been trying to fetch the metadata from a KDB+ Database using python, basically, I installed a library called qpython and using this library we connect and query the KDB+ Database.

I want to store the metadata for all the appropriate cols for a table/view in KDB+ Database using python. I am unable to separate the metadata part, despite trying myriad different approaches.

Namely a few to typecast the output to list/tuple, iterating using for , et cetera.

from qpython import qconnection

def fetch_metadata_from_kdb(params):
    try:
        kdb_connection_obj = qconnection.QConnection(host=params['host'], port=params['port'], username=params['username'], password=params['password'])
        kdb_connection_obj.open()
        PREDICATE = "meta[{}]".format(params['table'])
        metadata = kdb_connection_obj(PREDICATE)
        kdb_connection_obj.close()
        return metadata

    except Exception as error_msg: 
        return error_msg

def fetch_tables_from_kdb(params):
    try:
        kdb_connection_obj = qconnection.QConnection(host=params['host'], port=params['port'], username=params['username'], password=params['password'])
        kdb_connection_obj.open()
        tables = kdb_connection_obj("tables[]")
        views = kdb_connection_obj("views[]")
        kdb_connection_obj.close()
        return [table.decode() for table in list(tables)], [view.decode() for view in list(views)]

    except Exception as error_msg:
        return error_msg

parms_q = {'host':'localhost', 'port':5010,
           'username':'kdb', 'password':'kdb', 'table':'testing'}

print("fetch_tables_from_kdb:", fetch_tables_from_kdb(parms_q), "\n")
print("fetch_metadata_from_kdb:", fetch_metadata_from_kdb(parms_q), "\n")

The output which I am currently getting is as follows;

fetch_tables_from_kdb: (['testing'], ['viewname']) 

fetch_metadata_from_kdb: [(b'time',) (b'sym',) (b'price',) (b'qty',)]![(b'p', b'', b'') (b's', b'', b'') (b'f', b'', b'') (b'j', b'', b'')] 

I am not able to separate the columns part and the metadata part. How to store only the metadata for the appropriate column for a table/view in KDB using python?

Answer 1

The metadata that you have returned from kdb is correct but is being displayed in python as a kdb dictionary format which I agree is not very useful.

If you pass the pandas=True flag into your qconnection call then qPython will parse kdb datastructures, such as a table into pandas data structures or sensible python types, which in your case looks like it will be more useful.

Please see an example below - kdb setup (all on localhost)

$ q -p 5000
q)testing:([]date:.z.d+0 1 2;`g#sym:`abc`def`ghi;num:`s#10 20 30)
q)testing
date       sym num
------------------
2022.01.31 abc 10
2022.02.01 def 20
2022.02.02 ghi 30
q)meta testing
c   | t f a
----| -----
date| d
sym | s   g
num | j   s

Python code

from qpython import qconnection

#create and open 2 connections to kdb process - 1 without pandas flag and one
q = qconnection.QConnection(host="localhost", port=5000)
qpandas = qconnection.QConnection(host="localhost", port=5000, pandas=True)
q.open()
qpandas.open()

#see what is returned with a q table 
print(q("testing"))
[(8066, b'abc', 10) (8067, b'def', 20) (8068, b'ghi', 30)]
#the data is a qPython data object
type(q("testing"))
qpython.qcollection.QTable

#whereas using the pandas=True flag a dataframe is returned.
print(qpandas("testing"))
        date     sym  num
0 2022-01-31  b'abc'   10
1 2022-02-01  b'def'   20
2 2022-02-02  b'ghi'   30

#This is the same for the meta of a table
print(q("meta testing"))
[(b'date',) (b'sym',) (b'num',)]![(b'd', b'', b'') (b's', b'', b'g') (b'j', b'', b's')]

print(qpandas("meta testing"))
         t    f     a
c
b'date'  d  b''   b''
b'sym'   s  b''  b'g'
b'num'   j  b''  b's'

With the above you can now access the columns and rows using pandas (the b'num' etc is the qPython way of expressing a backtick `

Also now you have the ability to now use the DataFrame.info() to extract datatypes if you are more intrested in the python data structure rather than the kdb data structure/types. qPython will convert the q types to sensible python types automatically.

qpandas("testing").info()
<class 'pandas.core.frame.DataFrame'>
RangeIndex: 3 entries, 0 to 2
Data columns (total 3 columns):
 #   Column  Non-Null Count  Dtype
---  ------  --------------  -----
 0   date    3 non-null      datetime64[ns]
 1   sym     3 non-null      object
 2   num     3 non-null      int64
dtypes: datetime64[ns](1), int64(1), object(1)
memory usage: 200.0+ bytes

Answer 2

In the meantime, I have checked quite a bit of KBD documentation and found that the metadata provides the following as the output. You can see that here kdb metadata

c | tfa

c-columns t-symbol f-foreign key association a-attributes associated with the column

We can access the metadata object( <class 'qpython.qcollection.QKeyedTable'> ) by interating over a for loop as shown below;

from qpython import qconnection

def fetch_metadata_from_kdb(params):
    try:
        col_list, metadata_list = [], []
        kdb_connection_obj = qconnection.QConnection(host=params['host'], port=params['port'], username=params['username'], password=params['password'])
        kdb_connection_obj.open()
        PREDICATE = "meta[{}]".format(params['table'])

        ############# FOR LOOP ##############
        for i,j in kdb_connection_obj(PREDICATE).items():
            col_list.append(i[0].decode())
            metadata_list.append(j[0].decode())

        kdb_connection_obj.close()
        return col_list, metadata_list

    except Exception as error_msg: 
        return error_msg
    
parms_q = {'host':'localhost', 'port':5010,
           'username':'kdb', 'password':'kdb', 'table':'testing'}
           
print(fetch_metadata_from_kdb(parms_q))

Output: ['time', 'sym', 'price', 'qty'], ['p', 's', 'f', 'j']

I also got the KDB char types / q data types from the documentation here . Below is the implementation for the same;

import pandas as pd
from qpython import qconnection

kdb_type_char_dict = dict()

df = pd.read_html('https://code.kx.com/q4m3/2_Basic_Data_Types_Atoms/')[1].iloc[:17, 0:3][['Type', 'CharType']]
for i, j in zip(df.iloc[:, 0], df.iloc[:, 1]): kdb_type_char_dict[str(j)] = str(i)

####### Q DATA TYPES DICTIONARY #######
print("Chat types/ q data types dictionary:", kdb_type_char_dict)

def fetch_metadata_from_kdb(params):
    try:
        col_list, metadata_list, temp_list = [], [], []
        kdb_connection_obj = qconnection.QConnection(host=params['host'], port=params['port'],
                                                     username=params['username'], password=params['password'])
        kdb_connection_obj.open()
        PREDICATE = "meta[{}]".format(params['table'])
        for i, j in kdb_connection_obj(PREDICATE).items():
            col_list.append(i[0].decode())
            temp_list.append(j[0].decode())
        for i in temp_list:
            metadata_list.append("{}".format(kdb_type_char_dict[i]))
        kdb_connection_obj.close()
        return col_list, metadata_list

    except Exception as error_msg:
        return error_msg


params = {'host': 'localhost', 'port': 5010,
           'username': 'kdb', 'password': 'kdb', 'table': 'testing'}

print(fetch_metadata_from_kdb(params))

Output:

Chat types/ q data types dictionary: {'b': 'boolean', 'x': 'byte', 'h': 'short', 'i': 'int', 'j': 'long', 'e': 'real', 'f': 'float', 'c': 'char', 's': 'symbol', 'p': 'timestamp', 'm': 'month', 'd': 'date', 'z': '(datetime)', 'n': 'timespan', 'u': 'minute', 'v': 'second', 't': 'time'}


(['time', 'sym', 'price', 'qty'], ['timestamp', 'symbol', 'float', 'long'])