Shantanu's Blog

Machine learning is used to learn from given data and then predict values. For e.g. here is share price of a company for given years.

import numpy as np
xval = np.array([2001,2002,2003,2003,2004,2003,2006,2008,2009,2010]).reshape(-1,1)
yval = [1,2,3,4,5,6,7,7,9,10]

We need to create a model to store the data...

import sklearn.linear_model as skl
model = skl.LinearRegression()

The fit method of model will learn and help us predict values. In this case the price expected for the year 2012 is around 11.66

model.fit(xval,yval)
model.predict(2012)

array([ 11.66141732])

We can also plot the data to understand how the values are moving acorss years...

import pylab as py
py.scatter(xval,yval)

This is required if you get an error while initiating tokudb engine:

echo never > /sys/kernel/mm/transparent_hugepage/enabled

And this is required if you get permissions error:

rm -rf /storage/custom3381

mkdir /storage/custom3381

chown 1001 /storage/custom3381

percona server has built-in environment variable for tokudb:

docker run -p 3381:3306 -v /my/custom3381:/etc/mysql/conf.d -v /storage/custom3381:/var/lib/mysql -e MYSQL_ROOT_PASSWORD=india3381 -e INIT_TOKUDB=1 -d percona/percona-server:5.7

1) Download xtrabackup package
2) change directory
3) Full Backup
4) INcremental backup
5) Restore
6) Start mysql using backup
 
# Linux
wget https://www.percona.com/downloads/XtraBackup/Percona-XtraBackup-2.4.9/binary/tarball/percona-xtrabackup-2.4.9-Linux-x86_64.tar.gz

# centOS and redhat
yum install http://www.percona.com/downloads/percona-release/redhat/0.1-4/percona-release-0.1-4.noarch.rpm
yum install percona-xtrabackup-24
_____

cd percona-xtrabackup-2.4.9-Linux-x86_64

bin/xtrabackup --defaults-file=/my/custom3396/my.cnf -H 172.31.0.57 -uroot -pindia3396 -P 3396 --datadir /storage/mysql/datadir3396 --backup --target-dir=/data3/backups/full/
_____

The main advantage of using xtrabackup is that we can take incremental backup that will be much faster.

bin/xtrabackup --defaults-file=/my/custom3396/my.cnf -H 172.31.0.57 -uroot -pindia3396 -P 3396 --datadir /storage/mysql/datadir3396 --backup --target-dir=/data3/backups/inc1 --incremental-basedir=/data3/backups/full/


The next day, we need to simply change the target directory path to "inc2" like this:

bin/xtrabackup --defaults-file=/my/custom3396/my.cnf -H 172.31.0.57 -uroot -pindia3396 -P 3396 --datadir /storage/mysql/datadir3396 --backup --target-dir=/data3/backups/inc2 --incremental-basedir=/data3/backups/inc1
_____

In case of disaster we need to apply logs and then prepare data:

1) First apply logs of target directory:
bin/xtrabackup --prepare  --apply-log-only --target-dir=/data3/backups/full/

2) Apply logs from incremental backup:
bin/xtrabackup --prepare --apply-log-only --target-dir=/data3/backups/full/ --incremental-dir=/data3/backups/inc1

3) apply log only option should not be used for the last incremental backup.
bin/xtrabackup --prepare  --target-dir=/data3/backups/full/  --incremental-dir=/data3/backups/inc2

4) Finally prepare target without apply log option for target directory:
bin/xtrabackup --prepare --target-dir=/data3/backups/full/
_____

Now since the backup data directory is ready, we can create a new docker container pointing to the newly "prepared" data.

docker run -p 3391:3306 -e MYSQL_ROOT_PASSWORD=india3391 -v /my/custom3391:/etc/mysql/conf.d  -v /data3/backups/full:/var/lib/mysql -d shantanuo/mysql:5.7

You can check if the new data is working correctly.

mysql -h `hostname -i` -uroot -pindia3396 -P 3391

Here is how a standard class look like. When I call monthly function, I get the default 35000 value. I can however set a new value by calling another function called monthly_updated.

class pay_check:
    def __init__(self):
        self._salary = 35000

    def monthly(self):
        return self._salary
   
    def monthly_updated(self, value):
        self._salary=value

myclass=pay_check()
myclass.monthly()
myclass.monthly_updated(40000)
myclass.monthly()

This works, but it is possible to improve the usability of the class by adding property decorator. I make the monthly function as default getter that will be called when the user request the property method.

class pay_check:
    def __init__(self):
        self._salary = 35000

    @property
    def monthly(self):
        return self._salary

    @monthly.setter
    def monthly(self, value):
        self._salary=value

myclass=pay_check()

Instead of myclass.monthly() I can now simply use myclass.monthly (without brackets)
myclass.monthly

Another advantage is that I can use the same method to set the new value as shown below:
myclass.monthly=50000

Now the new value of salary is 50,000 as returned by this:
myclass.monthly

There are many advantages of using this style of programming. The code is readable, elegant and can be easily maintained. The user may slightly get confused with property concept since he has only seen functions as methods. But once he understand this, he can not live without it!

For e.g.
df.columns will return the column headings, but I can use the same function name to change the column names like this...
df.columns=['name', 'experience', 'remuneration', 'amount']

Or set a new value for the entire column:
df['dummy'] = '0'

And return the values of the given column using the same slice like this...
df['dummy']

Understanding how "get", "set" and "del" properties are handled in a class is very important to manage the class instances.

# Here is the python code that will check if any of the files in a given S3 bucket is publicly accessible. Change your-bucket-name, region and access / secret key

import boto
from boto.s3.connection import OrdinaryCallingFormat
conn = boto.s3.connect_to_region('ap-south-1', aws_access_key_id='xxx', aws_secret_access_key='xxx',calling_format=OrdinaryCallingFormat())

mybucket = conn.get_bucket('your-bucket-name')
for key in mybucket.list():
      for grant in key.get_acl().acl.grants :
            if grant.permission == 'READ' :
                print ("PUBLIC: " +str(key))
                #key.set_acl('private')

If you restart server or if docker ends abnormally like a

kill -9 {DOCKER_PID}

then you may get an error while restarting your containers.

# docker restart 2dc3fc6e5e3e d6d9d1dab040

Error response from daemon: Cannot restart container 2dc3fc6e5e3e: oci runtime error: container with id exists: 2dc3fc6e5e3e5b63c9d3ad8074972b72867b9ccd250b4c7fced42c616adc2070
Error response from daemon: Cannot restart container d6d9d1dab040: oci runtime error: container with id exists: d6d9d1dab0407706ef4ec37d0bacfe43134054ddd0b7a06d9b97434d0c288564

The solution is to remove containers from runc and containerd.
# rm -rf /run/runc/80768bc717f353484ab54b306bca0506861688d0b1ae0f3d724208cb37cad047
# rm -rf /run/containerd/80768bc717f353484ab54b306bca0506861688d0b1ae0f3d724208cb37cad047
# rm -rf /run/runc/2dc3fc6e5e3e5b63c9d3ad8074972b72867b9ccd250b4c7fced42c616adc2070
# rm -rf /run/containerd/2dc3fc6e5e3e5b63c9d3ad8074972b72867b9ccd250b4c7fced42c616adc2070

You can easily build ipython notebook environment using binder.

1) Visit binder page:
https://beta.mybinder.org

2) Type Github repo or URL:
https://github.com/psychemedia/showntell

3) Git branch:
maths

Click on launch. It will generate a ready-to-use environment that you can immediately start working on. select OpenLearn_Geometry.ipynb file and then select "show codecell inputs" button to show hidden cells.
_____

If you are using third-party modules in your code then you will need requirements.txt file in your repo with the names of all modules required to run your code. For e.g.

pandas
geopandas

If you need to execute certain commands after installing the modules, you will also need postBuild file. The contents of the file will look like this...

https://github.com/psychemedia/showntell/blob/maths/postBuild

Here is the code that will list all files in /home/ folder and create a nice data-frame.

import pandas as pd
from pathlib import Path
import time

p = Path(".")
all_files = []
for i in p.rglob('*.*'):
    all_files.append((i.name, i.parent, time.ctime(i.stat().st_ctime), i.stat()[6]))       

columns = ["File_Name", "Parent", "Created", "size"]
df = pd.DataFrame.from_records(all_files, columns=columns)

df.to_csv('file_list.csv', sep='\t')

It is easy to export it to excel, but I will prefer not to do that and continue working within ipython environment.

You can download and install chrome extension to download the history in json format for free.

#https://chrome.google.com/webstore/detail/hcohnnbbiggngobheobhdipbgmcbelhh/publish-accepted

The json file can be imported in pandas dataframe. You will need to change the epoch time to readable date-time and also find the domain names visited most.

import pandas as pd
from urllib.parse import urlparse
df=pd.read_json('history.json')
df['date'] = pd.to_datetime(df['lastVisitTime'],unit='ms' )

def extract(myurl):
    return urlparse(myurl).netloc

df['newurl']=df.url.apply(extract)
df.newurl.value_counts()