wip - i use this code to learn rust and kotlin

Song Unscrambler

Suppose you take an artist and song combination and scramble the letters. How would you retrieve the artist and song again without knowing that first step?

This application unscrambles that string into the artist and song.

How

This application builds an android application. The application contains an input field for scrambled data. The scrambled data is ordered and fed to query of MusicBrainz data. The corresponding artist and song are displayed back on the screen.

input

we can type the input by hand. Or use the Google Vision API to decode a photo to text.

database

Using the MusicBrainz database it creates all possible artist - song combinations. All letters of these values are then put in alphabetical order. This becomes the scrambled value.

If we now do the same with the unknown value we are able to look it up directly and derive the value.

Other implementation

Another implementation would use fuzzy search to find nearest neighbours.

Steps

application is in Android Kotlin, the PoC is done in rust

one time setup: creating the dataset

transform musicdb data to csv

The following steps show how to download the musicdb dataset, open it using docker, transform the data and export it to csv. We then have all artist song combinations.

download and extract database : ( this directory is not in git for (2G) of obvious reasons ... )

VERSION=$(curl -s https://mirrors.dotsrc.org/MusicBrainz/data/fullexport/LATEST)
curl -o mbdump.tar.bz2 -s https://mirrors.dotsrc.org/MusicBrainz/data/fullexport/${VERSION}/mbdump.tar.bz2
mkdir -p ${PWD}/musicdb/input
tar -xvjf mbdump.tar.bz2 -C ${PWD}/musicdb/input

download schema into mbdump/input: curl -o ${PWD}/musicdb/input/CreateTables.sql https://raw.githubusercontent.com/metabrainz/musicbrainz-server/master/admin/sql/CreateTables.sql

and the manualy remove the toc CUBE on file 3272 (and the comma on the line before)

enter the ${PWD}/musicdb directory
start a postgres container: docker run -ti -u postgres --name pg -v ${PWD}:/datadump postgres:9.5
in another shell enter the container as postgres docker exec -u postgres -ti pg bash
add user createuser musicbrainz_user
create db createdb -U postgres --owner=musicbrainz_user --encoding=UNICODE importtest
start psql
alter user with superuser rigts (needed to export csv) alter role musicbrainz_user with superuser;
exit \q
create tables and import tables (See https://wiki.musicbrainz.org/History:Database_Installation)

psql -U musicbrainz_user importtest

and import sql: \i /datadump/CreateTables.sql

exit: \q

import data:

for t in * ; do \
	echo `date` $t ; echo "\\copy $t from ./$t" | psql -U musicbrainz_user importtest ; \
done

enter psql again

psql -U musicbrainz_user importtest

select the data into temp table

  SELECT my.* 
  INTO tmp 
    FROM (
    select ac.name as artist, r.name as track 
    from recording r, artist_credit ac 
    where r.artist_credit = ac.id
  ) AS my;

export data to csv:

COPY tmp to '/tmp/tracks.csv' DELIMITER E'\t' QUOTE '"' CSV FORCE QUOTE *;

mv /tmp/tracks.csv /datadump/

the resulting file is about 800M

prepare sqlite database

Here use createDb.groovy script to read csv export, create a scrambled text and add it to an sqlite database.

To run this with test data:

head -1000 ${PWD}/musicdb/tracks.csv > tracks.csv
groovy ./createDb.groovy

The initial dataset contains 23.528.878 records. Or 14.121.143 records when using dropping all between brackets: (...)

However, for the first 10.000 records select count(*) from songs; yields 10.000 But select count(*) from (select distinct artist,song from songs); yields 68.000

So I replaced the id pk with a song,artist pk. And changed the INSERT into INSERT OR REPLACE

Together with the index on scramble slowed the creation of the database by a lot. However we now have a fast (and smaller) read-only database.

application

sanitize scrambled word.
find sanitized word in table
return song

Android App

Firebase and Google Cloud

https://cloud.google.com/vision/docs/quickstart

recognition

from android

https://cloud.google.com/vision/docs/reference/libraries#client-libraries-install-java

command line & testing

upload

Copy test file to google storage. This will be a photo from the phone's camera.

gsutil cp ~/Downloads/DSC_1419.JPG gs://song-unscrambler.appspot.com/test/DSC_1419.JPG
gsutil cp ~/Downloads/DSC_1421.JPG gs://song-unscrambler.appspot.com/test/DSC_1421.JPG

enable api

https://console.cloud.google.com/flows/enableapi?apiid=vision.googleapis.com&redirect=https:%2F%2Fconsole.cloud.google.com&pli=1&authuser=1

query

payload.json:

{
  "requests": [
    {
      "image": {
        "source": {
          "imageUri": "https://storage.googleapis.com/song-unscrambler.appspot.com/test/DSC_1421.JPG"
        }
      },
      "features": [
        {
          "type": "TEXT_DETECTION",
          "maxResults": 2
        }
      ]
    }
  ]
}

find key here:

https://console.cloud.google.com/apis/credentials?project=song-unscrambler&authuser=1

fire curl command:

API_KEY_VISION=my-key
curl -X POST \
    -H "Content-Type: application/json" -d @./scripts/vision-api-request.json \
    https://vision.googleapis.com/v1/images:annotate?key=$API_KEY_VISION > scripts/vision-api-response.json

add firebase

project is located here: https://console.firebase.google.com/u/1/project/song-unscrambler/settings/general/android:com.wijdemans.songunscrambler

download json

google-services.json can be downloaded

setup gradle

Project-level build.gradle (/build.gradle):

buildscript {
  dependencies {
    // Add this line
    classpath 'com.google.gms:google-services:3.1.0'
  }
}

App-level build.gradle (//build.gradle):

// Add to the bottom of the file
apply plugin: 'com.google.gms.google-services'

jeroenwijdemans / song-unscrambler