In theory, everything is organized.<\/p>\n
In practice, it quickly becomes difficult to find anything.<\/p>\n
The First Search Problem<\/h2>\n
With a handful of rocks, you can rely on memory.<\/p>\n
You remember roughly where things are.<\/p>\n
You recognize shapes and colors.<\/p>\n
But as the dataset grows, that breaks down.<\/p>\n
You start asking questions like:<\/p>\n
- \n
- Which rocks are under 5 pounds?<\/li>\n
- Which ones are suitable for landscaping?<\/li>\n
- Where did that smooth gray stone go?<\/li>\n<\/ul>\n
At that point, you\u2019re no longer dealing with a pile.<\/p>\n
You\u2019re dealing with a dataset.<\/p>\n
And datasets need to be searchable.<\/p>\n
Filtering by Metadata<\/h2>\n
The most straightforward approach is to use structured queries.<\/p>\n
If we have metadata like weight, color, and classification, we can filter directly.<\/p>\n
Conceptually:<\/p>\n
SELECT *\nFROM rocks\nWHERE weight_lb < 5\nAND color = 'gray'\nAND rock_class <= 'Class 2'\n<\/code><\/pre>\nThis works well for clearly defined attributes.<\/p>\n
It\u2019s predictable.<\/p>\n
It\u2019s efficient.<\/p>\n
And it\u2019s the foundation of most data systems.<\/p>\n
The Role of Classification<\/h2>\n
This is where the Quarry Taxonomy starts to pay off.<\/p>\n
Instead of requiring precise measurements, we can use categories:<\/p>\n
- \n
- Pebble Class<\/li>\n
- Hand Sample<\/li>\n
- Landscaping Rock<\/li>\n
- Wheelbarrow Class<\/li>\n
- Engine Block Class<\/li>\n<\/ul>\n
This allows for simpler queries:<\/p>\n
- \n
- \u201cShow me everything below Wheelbarrow Class\u201d<\/em><\/li>\n
- \u201cExclude Engine Block Class entirely\u201d<\/em><\/li>\n<\/ul>\n
Classification reduces complexity.<\/p>\n
It turns continuous values into discrete groups.<\/p>\n
This is a common pattern in real-world systems.<\/p>\n
When Metadata Isn\u2019t Enough<\/h2>\n
Structured queries work well when you know exactly what you\u2019re looking for.<\/p>\n
But sometimes you don\u2019t.<\/p>\n
Sometimes the question looks more like:<\/p>\n
Find rocks that look like this one.<\/em><\/p>\n
Or:<\/p>\n
Find something similar to the smooth stone I saw earlier.<\/em><\/p>\n
At that point, metadata alone isn\u2019t enough.<\/p>\n
We need another way to compare objects.<\/p>\n
Similarity and Representation<\/h2>\n
Images and 3D models contain information that isn\u2019t captured in simple fields like color or weight.<\/p>\n
To use that information, we need to represent it in a comparable way.<\/p>\n
One approach is to generate embeddings \u2014 numerical representations of images or shapes.<\/p>\n
Conceptually:<\/p>\n
- \n
- each rock image \u2192 vector representation<\/li>\n
- similar images \u2192 vectors close together<\/li>\n
- dissimilar images \u2192 vectors further apart<\/li>\n<\/ul>\n
This allows for similarity search.<\/p>\n
Instead of filtering by attributes, we search by resemblance.<\/p>\n
A Different Kind of Query<\/h2>\n
With similarity search, queries look different.<\/p>\n
Instead of:<\/p>\n
color = 'gray'\nweight < 5\n<\/code><\/pre>\nWe might have:<\/p>\n
find nearest neighbors to this image\n<\/code><\/pre>\nThis shifts the system from exact matching to approximate matching.<\/p>\n
It\u2019s less precise.<\/p>\n
But often more useful.<\/p>\n
A Familiar Pattern<\/h2>\n
At this point, the Backyard Quarry<\/strong> starts to resemble systems used in:<\/p>\n
- \n
- image search engines<\/li>\n
- product recommendation systems<\/li>\n
- digital asset management platforms<\/li>\n
- AI-powered retrieval systems<\/li>\n<\/ul>\n
The objects are different.<\/p>\n
The pattern is the same.<\/p>\n
Store data.<\/p>\n
Index it.<\/p>\n
Provide multiple ways to retrieve it.<\/p>\n
Combining Approaches<\/h2>\n
In practice, the most useful systems combine both methods.<\/p>\n
Structured filtering:<\/p>\n
- \n
- weight<\/li>\n
- class<\/li>\n
- location<\/li>\n<\/ul>\n
Similarity search:<\/p>\n
- \n
- appearance<\/li>\n
- shape<\/li>\n
- texture<\/li>\n<\/ul>\n
Together, they provide flexibility.<\/p>\n
You can narrow down the dataset and then explore it.<\/p>\n
The Cost of Search<\/h2>\n
Search doesn\u2019t come for free.<\/p>\n
It introduces:<\/p>\n
- \n
- indexing overhead<\/li>\n
- additional storage<\/li>\n
- preprocessing steps<\/li>\n
- more complex queries<\/li>\n<\/ul>\n
And like everything else in the Quarry system, these tradeoffs become more significant as the dataset grows.<\/p>\n
The Realization<\/h2>\n
At this point, something interesting becomes clear.<\/p>\n
The hard part isn\u2019t collecting rocks.<\/p>\n
It isn\u2019t even modeling them.<\/p>\n
The hard part is making the data usable.<\/p>\n
And usability, in most systems, comes down to one thing:<\/p>\n
Search.<\/p>\n
What Comes Next<\/h2>\n
With data captured and searchable, the next step is to zoom out.<\/p>\n
What we\u2019ve built so far is more than just a rock catalog.<\/p>\n
It\u2019s a small example of a larger idea.<\/p>\n
In the next post, we\u2019ll look at that idea more directly:<\/p>\n
Digital twins.<\/p>\n
Because once you can represent, store, and search objects, you\u2019ve taken the first step toward building systems that mirror the physical world.<\/p>\n
And somewhere in the process, it becomes clear that even a pile of rocks benefits from thoughtful indexing.<\/p>\n
Which is not something I expected to say when this started.<\/p>\n
The Rock Quarry Series<\/h2>\n
- \n
- The Backyard Quarry: Turning Rocks Into Data<\/a><\/li>\n
- The Backyard Quarry, Part 2: Designing a Schema for Physical Objects<\/a><\/li>\n
- The Backyard Quarry, Part 3: Capturing the Physical World<\/a><\/li>\n<\/ul>\n
![\"Facebook\"](\"https:\/\/i0.wp.com\/www.kenwalger.com\/blog\/wp-content\/plugins\/social-media-feather\/synved-social\/image\/social\/regular\/96x96\/facebook.png?resize=48%2C48&ssl=1\" "\\"Share")
<\/a>![\"twitter\"](\"https:\/\/i0.wp.com\/www.kenwalger.com\/blog\/wp-content\/plugins\/social-media-feather\/synved-social\/image\/social\/regular\/96x96\/twitter.png?resize=48%2C48&ssl=1\" "\\"Share")
<\/a>![\"reddit\"](\"https:\/\/i0.wp.com\/www.kenwalger.com\/blog\/wp-content\/plugins\/social-media-feather\/synved-social\/image\/social\/regular\/96x96\/reddit.png?resize=48%2C48&ssl=1\" "\\"Share")
<\/a>![\"linkedin\"](\"https:\/\/i0.wp.com\/www.kenwalger.com\/blog\/wp-content\/plugins\/social-media-feather\/synved-social\/image\/social\/regular\/96x96\/linkedin.png?resize=48%2C48&ssl=1\" "\\"Share")
<\/a>![\"mail\"](\"https:\/\/i0.wp.com\/www.kenwalger.com\/blog\/wp-content\/plugins\/social-media-feather\/synved-social\/image\/social\/regular\/96x96\/mail.png?resize=48%2C48&ssl=1\" "\\"Share")
<\/a>","protected":false},"excerpt":{"rendered":"By this point, the Backyard Quarry has a schema, a capture process, and a growing collection of records. Each rock has: metadata images possibly a 3D model In theory, everything is organized. In practice, it quickly becomes difficult to find anything. The First Search Problem With a handful of rocks, you can rely on memory. … Continue reading “The Backyard Quarry, Part 4: Searching a Pile of Rocks”<\/span><\/a><\/p>\n","protected":false},"author":1,"featured_media":0,"comment_status":"open","ping_status":"open","sticky":false,"template":"","format":"standard","meta":{"pmpro_default_level":"","_monsterinsights_skip_tracking":false,"_monsterinsights_sitenote_active":false,"_monsterinsights_sitenote_note":"","_monsterinsights_sitenote_category":0,"jetpack_post_was_ever_published":false,"_jetpack_newsletter_access":"","_jetpack_dont_email_post_to_subs":false,"_jetpack_newsletter_tier_id":0,"_jetpack_memberships_contains_paywalled_content":false,"_jetpack_memberships_contains_paid_content":false,"footnotes":""},"categories":[1739,1738],"tags":[1736,1713],"yst_prominent_words":[810],"class_list":["post-1347","post","type-post","status-publish","format-standard","hentry","category-data-engineering","category-software-engineering","tag-data-engineering","tag-system-design","pmpro-has-access"],"jetpack_featured_media_url":"","jetpack_sharing_enabled":true,"jetpack_shortlink":"https:\/\/wp.me\/p8lx70-lJ","jetpack-related-posts":[],"_links":{"self":[{"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/posts\/1347","targetHints":{"allow":["GET"]}}],"collection":[{"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/posts"}],"about":[{"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/types\/post"}],"author":[{"embeddable":true,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/users\/1"}],"replies":[{"embeddable":true,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/comments?post=1347"}],"version-history":[{"count":2,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/posts\/1347\/revisions"}],"predecessor-version":[{"id":1349,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/posts\/1347\/revisions\/1349"}],"wp:attachment":[{"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/media?parent=1347"}],"wp:term":[{"taxonomy":"category","embeddable":true,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/categories?post=1347"},{"taxonomy":"post_tag","embeddable":true,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/tags?post=1347"},{"taxonomy":"yst_prominent_words","embeddable":true,"href":"https:\/\/www.kenwalger.com\/blog\/wp-json\/wp\/v2\/yst_prominent_words?post=1347"}],"curies":[{"name":"wp","href":"https:\/\/api.w.org\/{rel}","templated":true}]}}
- The Backyard Quarry, Part 2: Designing a Schema for Physical Objects<\/a><\/li>\n
- The Backyard Quarry: Turning Rocks Into Data<\/a><\/li>\n
- \u201cExclude Engine Block Class entirely\u201d<\/em><\/li>\n<\/ul>\n
- \u201cShow me everything below Wheelbarrow Class\u201d<\/em><\/li>\n