Before runnning the following commands, make sure that the docker container is running, and that the database is populated. Run the following commands to access the HBase shell:
docker exec -it hadoop bash
cd /usr/local/hbase/bin/
./hbase shell
After that import this class to use the Bytes.toBytesBinary method:
import org.apache.hadoop.hbase.util.Bytes
- Get the user (ssample4) info to check if a the username and password are correct
get 'user', 'ssample4', 'info'
- Create a new user with the username password name and email (ssample10, , 1234, sample,[email protected]) and get the new user info
put 'user', 'ssample10', 'info:name', 'sample'
put 'user', 'ssample10', 'info:password', '$2b$12$uqT6QX.6t.5cZvke1iVVTu3iT7VEGZMG2dWNA9m.qE5NA.TvBCVly'
put 'user', 'ssample10', 'info:email', '[email protected]'
get 'user', 'ssample10', 'info'
- Get the most recent price of the stock (AAPL)
NOTE: There is in fact a bug in the hbase shell that is not possible to use the reverse order in the scan command when a row prefix is used. The solution is to use StartRow and EndRow. The bug is documented in the following link: https://issues.apache.org/jira/browse/HBASE-16886?jql=project%20%3D%20HBASE. It appears that the bug was fixed as the status is Closed but it is still present.
scan 'instrument_prices', {STARTROW => 'AAPL`', ENDROW => 'AAPL_', LIMIT => 1, REVERSED => true}
- Retrieve historical prices and variation of an item across various time intervals (last week, month, year) with different levels of detail (per minute, hour, day)
This is actually not possible to do in the hbase shell. Is necessary to use the MapReduce witch is not possible to use in the hbase shell. The solution is to use the Hadoop MapReduce.
- Price updates for a stock (AAPL)
put 'instrument_prices', Bytes.toBytesBinary("AAPL_\x00\x00\x01\xb8\xda\xc5\xb4\x00"), 'series:val', '100'
scan 'instrument_prices', {STARTROW => 'AAPL`', ENDROW => 'AAPL_', LIMIT => 1, REVERSED => true}
- See all the posts for a given a financial instrument (AAPL)
scan 'financial_instruments', {ROWPREFIXFILTER => 'AAPL', COLUMNS => 'posts'}
- See all the posts of a given user (ssample4)
scan 'user', {ROWPREFIXFILTER => 'ssample4', COLUMNS => 'posts'}
- Create a new post for a given user (ssample4) and financial instrument (AAPL)
put 'user', 'ssample4', "posts:\x7f\xff\xfem\x87V\xea\x7f", "{\"symbol\":\"AAPL\",\"post\":\"This is a new post.\",\"post_id\":\"100000\"}"
put 'financial_instruments', 'AAPL', "posts:\x7f\xff\xfem\x87V\xea\x7f", "{\"username\":\"ssample4\",\"post\":\"This is a new post.\", \"post_id\":\"100000\"}"
- See all the followers/following of a given user (ssample4)
get 'user', 'ssample4', {COLUMN => 'followers'}
get 'user', 'ssample4', {COLUMN => 'following'}
- Follow a user (ssample4) from another user (ssample10)
put 'user', 'ssample4', 'followers:ssample10', '1'
incr 'user', 'ssample4', 'info:followers', 1
incr 'user', 'ssample10', 'info:following', 1
put 'user', 'ssample10', 'following:ssample4', '1'
*Get the most popular financial instruments
scan 'popularity_to_instrument', {LIMIT => 10}
-
Descriptive statistics for a given item like Day's Range, 52 Week Range, Mean, etc. This is actually not possible to do in the hbase shell. Is necessary to use the MapReduce witch is not possible to use in the hbase shell. The solution is to use the Hadoop MapReduce.
-
Look up the financial instrument (AAPL) by its symbol
get 'financial_instruments', 'AAPL'
- Retrieve a user's portfolio (ssample4).
scan 'portfolio', { FILTER => "PrefixFilter('ssample4_')" }
- See a user's trading history.
scan 'user', { STARTROW => 'ssample4', ENDROW => 'ssample4+', COLUMN => 'trades'}
-
Buy a stock (AAPL) for a user (ssample4) and update the user's portfolio. This is actually not possible to do in the hbase shell. Is necessary the use of transactions witch is not possible to use in the hbase shell.
-
Add funds (10$) to a user's account (ssample4) and update the user's balance. Notice that the value is multiplied by 100 because the hbase counter are integers, so the value is multiplied by 100 to store the cents, and then in the application, the value is divided by 100 to get the dollars.
incr 'user', 'ssample4', 'info:balance', 1000