Best way to store and search keywords for a record in PHP and MySQL?

Question

I haven't touched any code in a good 4-5 months so just getting back into it today, usually takes me a week or so to get all the info flowing through my brain again once I take months off like that. So my project I am about to start will be a PHP/MySQL backend bookmarks database.

I want to create a nice searchable database with all my favorite websites/bookmarks. Each record will have multiple keywords assigned to it so I can easily search all my bookmarks for the term "php" and all records with "php" in there keyword column or title or otherwise will come back in a result set.

Here is my idea for the database so far...

auto_id = Auto incremented ID number for database
name/title = Name/title of the Website
description = brief description of the site
URL = URL to open when I click a link
clicks = increments by 1 everytime I click the link
date_created = datetime that URL bookmark was added
date_accessed   = datetime field for when last clicked on
category = category name or number to create a folder like structure of bookmarks in groups
sub_category = some categories will have subcategories (ie programming->c##  programming->PHP )
keywords = Keywords used for searching

This is pretty straight forward for me on how to build this system all except I am looking for help/advice on the best way to store the keywords. Each website/record I add to the DB can have 1 up to multiple keywords per site. These keywords need to be able to help with the searching part of my app. So how should I store keywords for a site in my database? I know I could just have a "keywords" row in the table and store the keywords for each record like this "php, web, etc, keyword4" so all keywords for each site are saved in 1 column but this does not seem to be the best method when it comes to searching the database.

Please tell me how you would do this part? Thanks for any help

Answer 1

The easiest, and fastest, search technique to implement is the use of MySQL's LIKE statement. LIKE lets you search through a column for a specific string. Consider the following example...

auto_id    name            description
1          Cool PHP Site   you know you love it  
2          PLARP! its Ruby gems gems gems!  
3          SqlWha          sql for the masses  
4          FuzzD00dle      fun in the sun, with some fuzz  

You could find all rows that contain the string 'php' in either the 'name' or 'description' field using the following query...

SELECT * FROM bookmarks WHERE name LIKE '%php%' OR description LIKE '%php%';

• '%' is a wildcard character.

Reference on MySQL LIKE: http://www.tutorialspoint.com/mysql/mysql-like-clause.htm

You could also add a 'keywords' column and store the keywords in a comma delimited format (ie: plarp1, plarp2, plarp3), then search through that.

Answer 2

You're right, storing a comma-separated list in one column is not a good way to do it (this is called a repeating group and it violates the First Normal Form of relational database design).

Using a LIKE predicate is not a good choice, because it cannot benefit from an index. Searching for keywords this way is hundreds or thousands of times slower than designing a proper database in normal form, and adding indexes.

You need to store a second table listing keywords, and a third many-to-many table to pair keywords to applicable bookmarks. This is a pretty standard design for "tagging" in a relational database.

In non-relational databases like CouchDB or MongoDB, you can make one field a set of keywords, and index them so queries can be efficient. But not in a relational database.

See also:

• How do you recommend implementing tags or tagging
• Database Design for Tagging

Also when viewing those questions, check the many related questions in the column on the right.

Answer 3

The best way to do this is to create a separate table to contain your keywords and then add an intersection (or join) table to join keywords with bookmarks.

CREATE TABLE bookmarks (
  id INT NOT NULL,
  ... etc.
)

CREATE TABLE keywords (
  id INT NOT NULL,
  ... etc.
)

CREATE TABLE bookmark_keywords (
  bookmark_id INT NOT NULL,
  keyword_id INT NOT NULL,
  PRIMARY KEY (bookmark_id, keyword_id),
  FOREIGN KEY bookmark_id REFERENCES bookmarks (id),
  FOREIGN KEY keyword_id REFERENCES keywords (id)
)

When you insert a bookmark, you'd also insert any keywords that are being used and aren't already in the keywords table, as well as a row in bookmark_keywords in order to join the keyword with the bookmark.

Then, when you want to query for what keywords a bookmark has:

SELECT k.*
FROM keywords AS k
LEFT JOIN bookmark_keywords AS kb
  ON kb.keyword_id = k.id
WHERE kb.bookmark_id = [ID of the bookmark]

And to query for what bookmarks share a particular keyword:

SELECT b.*
FROM bookmarks AS b
LEFT JOIN bookmark_keywords AS kb
  ON kb.bookmark_id = b.id
WHERE kb.keyword_id = [ID of the keyword]