This Project creates a web server (listening on port 8080) with aGET /search endpoint that will return the most appropriate 20 items given searchTerm, lat (latitude) and lng (longitude).
e.g. /search?searchTerm=camera&lat=51.948&lng=0.172943
go version 1.8.1 and higher
Run go get github.com/ObsidianRock/backend-challenge and go to the new directory and run go install
This will put the executable in the directory $GOPATH/bin ( this should be in your system environment variables)
From there onward, you can simple run the backend-challenge.exe and will spin up a server listening on port 8080
Download the executable from the release section on the GitHub project
On windows
# github.com/mattn/go-sqlite3
exec: "gcc": executable file not found in %PATH%gccThe project uses go-sqlite3 as SQLite driver. go-sqlite3 uses cgo hence need a C compiler.
On windows, Install TDM-GCC.
go the directory $GOPATH/src/ObsidianRock/backend-challenge and run the command:
go test ./... to run all tests in current directory and all of its subdirectories.
There are four main packages in the project,
The Tokenizer function takes a string input and returns a slice of discrete words from the input string, the function will also remove any non-alphanumeric characters and lower case the word:
For example
sentence = "Canon EF 24-70mm "
tokens = ["canon", "ef", "2470mm"]
refer to package test for more examples.
This package takes user input from URL query to create the SQL query.
The query parameters is checked against a map of predefined search terms (searchTerm, lat andlng), anything else raises an error. The search terms is tokenized and a SQL prepared statement is created dynamically based on the number of token words. The prepared statement is used to prevent SQL injection.
example of generated query:
SELECT * FROM items WHERE item_name LIKE ? or item_name LIKE ? or etc..
The SQL LIKE operator is used to search for a specified pattern in a column.
In this case it will be LIKE %token_word% to find any value that that have "token_word" in any position
See the package tests for examples.
The weigh package will take item and give it an importance based on two factors; distance and rank.
The distance in meters between the item and the search lat and lng is calculated using the haversine formula. For this project, the open source project tile38 has a good function which implements this formula and used in this project. The rank determines the relevance of an item based on the number of time an item name matches the tokenized search terms.
Take a look at the package tests for examples.
The query package will execute the SQL query with the parameters, due to the use of the SQL LIKE parameter, the token words needs to be pre and post appended with the symbol % .
The distance and rank of each item will also be calculated.
To keep it simple, the result is sorted based on decreasing rank and increasing distance.
See the package tests for examples.
To keep it simple, the project omits pagination. To include pagination, a suggestion would to track some form of metadata in the form of offset and limit, so the output JSON will include those and next query will include the metadata to get the next two most prominent results.
According to this report, using prepared statement could lead to performance degrading due fact that each query is generating a new prepared statement. The improvement would be to remove the prepared statement as the database/sql query function creates prepares statements under the hood according to this guide.
SQLite a fantastic database for testing a prototyping, however using a more robust database such a PostgreSQL for production would be better as well as the fact that it can be extended to include spatial database support with PostGIS is awesome.