AmberDB

🏠 Home  •  📖 About  •  🚀 Quick Start  •  📘 Tutorial  •  🌐 Locale  •  📋 Changes  •  📚 Wiki  •  🇹🇷 Türkçe

Developer Guide and Comprehensive Documentation

Version: 5.23.0 · Initial Design: 2005 · Last Updated: 2026
Namespace: AmberDB
Built-in Modules: Base, Index, Transact, Cache, Array, String, Date, Locale, Tools

Table of Contents

  1. What is AmberDB?
  2. Quick Start
  3. CRUD Operations (Core Data Management)
  4. Reading, Filtering, and Sorting
  5. Simple Mode and Direct Schemaless Access (Simple Mode)
  6. Indexing and Search Engine
  7. Transaction Safety, ACID Guarantees, and Crash Recovery (Transactions)
  8. High-Throughput Batch Operations (Batch ETL & Ingestion)
  9. Schema Configuration (.table & In-Memory)
  10. Database Group Structure (.dbase)
  11. Smart Tiered (Hot / Cold Junk) Indexing
  12. Automated URL Slug Management
  13. Unified Shared RAM Cache (.db / .inx) & Persistent Buffer
  14. Configuration and Deterministic Flag Management (config)
  15. Data Structures, Low-Level Table and Stream Operations
  16. Faceted Search & Category Filters (Facet Engine)
  17. User Audit Trail and Backup
  18. Maintenance and Repair Tools (AmberDB::Tools)
  19. File Extensions Map
  20. Directory Structure
  21. Developer Best Practices and Recommendations
  22. Full Working Example (Checkout & Stock Transaction Scenario)
  23. Method Quick Reference Table
  24. Why Use AmberDB? (Comparison with SQL and SQLite)
  25. Boundaries and Debated Topics (Physical Constraints vs. Conscious Architectural Choices)

1. What is AmberDB?

AmberDB is a high-performance, schema-driven NoSQL database engine for Perl, featuring precomputed inverted indexing, ACID-compliant transactions with Strict Two-Phase Locking (Strict 2PL), and automatic crash recovery on top of Berkeley DB (DB_File).

From a developer’s perspective, AmberDB eliminates the overhead of provisioning and maintaining external database servers. A single CRUD call automatically updates and synchronizes all associated full-text search, field-match, facet filter, binary sort, and bidirectional URL slug indexes in one integrated layer.

Built-in Modular Architecture

AmberDB is self-contained and does not rely on heavy external dependencies:

┌─────────────────────────────────────────────────────────────────────────┐
│                              AmberDB                                    │
├─────────────────────────────────────────────────────────────────────────┤
│  AmberDB::Base     → Schema parsing, paths, data serialization          │
│  AmberDB::Index    → Binary indexes (.inx, .fld, .src, .fac, .srt)      │
│  AmberDB::Transact → Undo-log transactions, rollback & recovery         │
│  AmberDB::Cache    → Native RAM-Disk (tmpfs) Shared Cache & TTL         │
│  AmberDB::Array    → High-speed array utilities (nodup, crop)           │
│  AmberDB::String   → String utilities, HTML formatting & cleaning       │
│  AmberDB::Date     → Date calculations, timestamps, formatting          │
│  AmberDB::Locale   → Built-in multilingual collation & word search      │
├─────────────────────────────────────────────────────────────────────────┤
│  AmberDB::Tools    → Standalone reindexing, vacuum & repair tools       │
└─────────────────────────────────────────────────────────────────────────┘

Note: Collation-aware multilingual sorting and searching are powered by the integrated AmberDB::Locale module and require no external services or third-party packages.

2. Quick Start

2.1 Instantiating the Database Object

use AmberDB;

my $adb = AmberDB->new(
    cfg  => { 
        language => "en",          # Built-in Locale language ("en", "tr", "de" etc.)
    },
    path => { 
        dbase_dir => "./dbstore",  # Database root directory
    },
);

[!TIP] Variable Naming Convention ($adb): Throughout AmberDB documentation and code examples, the variable name $adb (AmberDB Handle) is used to represent the database instance, following Perl’s standard $dbh convention. While not mandatory, using $adb is recommended to maintain clean code readability and prevent namespace collisions in hybrid architectures that concurrently use relational DBI $dbh and AmberDB.

Database Directory Convention (dbstore): Similarly, dbstore is used throughout the documentation and examples as the canonical database root directory name. This is merely a standard convention and not a hardcoded requirement; you may designate any folder name or directory path (e.g., data, db, storage, /var/data/myapp, etc.) that best fits your environment. However, for cross-platform filesystem consistency, the directory path must consist strictly of lowercase ASCII characters (avoiding spaces, uppercase, or non-ASCII characters).

2.2 Directory Configuration

AmberDB automatically configures the required subdirectories under the base path upon initialization. If you need to assign or modify the database directory after object creation, use the set_datadir() method:

# Dynamically change data directory if needed
$adb->set_datadir("/var/data/myapp/other/dir");

[!WARNING] Security Notice (Restrict Web Access): Ensure that the root database directory (dbase_dir) assigned during AmberDB initialization is located outside the web server’s public document root (public_html, htdocs, www, etc.) or is strictly shielded by web server configuration rules (.htaccess, Nginx block rules) so that database files cannot be accessed or downloaded via public HTTP requests.

2.3 Structure and Anatomy of an AmberDB Record

In AmberDB, every record (document) is natively represented as a Perl list/array structure (@record). Unlike rigid relational columns in SQL, AmberDB records are lightweight, highly flexible, and object-oriented:

[!TIP] Essential AmberDB Core Methods:
The primary methods used in everyday application workflows are:

# =========================================================================
# 1. Constructing and Inserting a Record (@record)
# =========================================================================
# In AmberDB, the recommended practice is to maintain the Record ID at Index 0
# of the array ($record[0] = 0 for new records) and manage the array holistically:
my @record = (
    0,                                  # [0] Index: Record ID (0 or undef: Auto-Increment)
    "John Doe",                         # [1] Index: Full Name (Scalar Text)
    "john.doe@example.com",             # [2] Index: Email (Scalar Text)
    "5,12",                             # [3] Index: Category IDs (Relational List)
    1249.90,                            # [4] Index: Balance / Amount (Numeric)
    [ "Role_Admin", "Role_Editor" ],    # [5] Index: Permissions (Nested ARRAY reference)
    { status => "active", login_count => 12 }, # [6] Index: Metadata (Nested HASH reference)
);

# Insert: Assign generated ID to both variable and array index 0:
my $id = $record[0] = $adb->insert_id("member_user", @record);
print "Record created successfully with ID: $id\n";

# =========================================================================
# 2. Reading, Updating, and Deleting (Standard CRUD Lifecycle)
# =========================================================================
# Read: Returned array contains the Primary Key at Index 0:
my @retrieved = $adb->read_id("member_user", $id);

my $record_id   = $retrieved[0]; # Equals $id (e.g. 1001)
my $name        = $retrieved[1]; # "John Doe"
my $email       = $retrieved[2]; # "john.doe@example.com"
my $permissions = $retrieved[5]; # [ "Role_Admin", "Role_Editor" ] (ARRAY-ref)
my $metadata    = $retrieved[6]; # { status => "active", ... } (HASH-ref)

# Update: Modify fields and pass @retrieved directly to modify_id:
$retrieved[4] = 1499.90; # Update balance
$adb->modify_id("member_user", @retrieved);

# Delete: Remove record using Index 0 ID:
$adb->delete_id("member_user", $retrieved[0]);

3. CRUD Operations (Core Data Management)

In AmberDB, core insertion, modification, deletion, and retrieval operations are executed directly against the database table.

Creating a .table schema file is not strictly mandatory; schemaless tables store and retrieve records by primary key with zero configuration. However, if indexing directives are defined in the table schema (record_index, match_block, search_block, facet_block, sort_block, slug_block):

  1. Every insert_id, modify_id, or delete_id call automatically compiles, synchronizes, and maintains all secondary search, match, and sort indexes in the background.
  2. Read, query, and search operations (read_all, field_fetch, search_table, facet_menu, etc.) automatically utilize these precomputed indexes, bypassing slow full disk scans and executing via direct index lookups.

3.1 Inserting Records — insert_id

In AmberDB, relational fields (configured via match_block and rdbm) store foreign primary keys (IDs) rather than plain text strings.

# =========================================================================
# STEP 1: Populate Master Entity Tables
# =========================================================================
# 1. Category Table (catalog_category):
my $cat_computers = $adb->insert_id("catalog_category", 0, "Computers & IT", 1); # ID: 5
my $cat_audio     = $adb->insert_id("catalog_category", 0, "Headphones & Audio", 1); # ID: 12

# 2. Producer / Brand Table (catalog_brand):
my $brand_sony    = $adb->insert_id("catalog_brand", 0, "Sony", "Japan");          # ID: 3
my $brand_apple   = $adb->insert_id("catalog_brand", 0, "Apple", "USA");           # ID: 8

# 3. Contributor / Author Table (catalog_author):
my $author_1      = $adb->insert_id("catalog_author", 0, "John Doe", "Audio Eng"); # ID: 7
my $author_2      = $adb->insert_id("catalog_author", 0, "Jane Smith", "Designer");# ID: 9

# =========================================================================
# STEP 2: Inserting a Product Record (catalog_product)
# =========================================================================
# IMPORTANT:
# - Blocks 1 (Category), 2 (Brand), and 3 (Author) must be passed as IDs
#   belonging to their respective master tables, NOT raw text.
# - Multi-category or multi-author assignments are concatenated with commas ("5,12" or "7,9").
# - Standard Practice: Set index 0 to 0 and pass the whole array to insert_id.

my @product_data = (
    0,                            # [0] Record ID (0: Auto-Increment ID)
    "5,12",                       # [1] Category IDs (Multi-value: 5 = Computers, 12 = Audio)
    "3",                          # [2] Brand ID (3 = Sony)
    "7,9",                        # [3] Author / Contributor IDs (Multi-value: Authors 7 and 9)
    "WH-1000XM5 Wireless Headphones", # [4] Title
    "Active Noise Cancelling ANC",# [5] Subtitle
    "Supplier Inc.",              # [6] Supplier
    "Sony WH-1000XM5 premium sound...", # [7] Description
    "",                           # [8] Extra specs
    "8690001234567",              # [9] Barcode
    "399.90",                     # [10] Price
    "1"                           # [11] Status (1: Active)
);

# Inserting with auto-generated ID (Assigned ID updates both variable and $product_data[0])
my $new_id = $product_data[0] = $adb->insert_id("catalog_product", @product_data);
print "Inserted product ID: $new_id\n";

# Inserting with an explicit custom Primary Key ID (Assign custom ID to index 0):
$product_data[0] = 5001;
$adb->insert_id("catalog_product", @product_data);

# =========================================================================
# STEP 3: How Multi-Value Lookups (field_fetch) Work
# =========================================================================
# AmberDB's 'field_to_list' feature automatically unpacks comma-delimited strings
# ("5,12" and "7,9") and indexes each discrete ID into its respective .fld index.
# Both of the following independent queries will immediately find the product via fast direct index lookup:
my @cat12_items   = $adb->field_fetch("catalog_product", 1, "12"); # All products in Category 12
my @author9_items = $adb->field_fetch("catalog_product", 3, "9");  # All products by Author 9

3.2 Updating Records — modify_id

In AmberDB, updating records is performed consistently and holistically using the record array (@record or @fields) where Index 0 contains the target Record ID. The modify_id method automatically consumes the first element ($record[0]) as the Primary Key ID:

# Approach 1: Read existing record via read_id, update fields, and save
my @record = $adb->read_id("catalog_product", 5001);

$record[1]  = "5,12,18";   # Add category 18
$record[10] = "429.90";    # Update price

my $ok = $adb->modify_id("catalog_product", @record);

# Approach 2: Constructing an update array with data from a Form/API
my $record_id = 5001; # Target ID incoming from a web form, URL, or API payload

my @fields = (
    $record_id,                   # [0] Target Record ID (Variable or scalar)
    "5,12,18",                    # [1] Category IDs
    "3",                          # [2] Brand ID
    "7,9",                        # [3] Author IDs
    "WH-1000XM5 Headphones",      # [4] Product Title
    "Updated Description",        # [5] Subtitle
    "Supplier Inc.",              # [6] Supplier
    "Detailed sound...",          # [7] Description
    "",                           # [8] Specs
    "8690001234567",              # [9] Barcode
    "429.90",                     # [10] Price
    "1"                           # [11] Status
);

my $ok2 = $adb->modify_id("catalog_product", @fields);

if ($ok || $ok2) {
    print "Product and all related indexes updated successfully.\n";
}

[!WARNING] Since the record array (@record / @fields) already contains the Record ID at Index 0, do not pass an extra ID argument after the table name (i.e. avoid $adb->modify_id("table", 5001, @fields)). Always pass the array directly.

3.3 Deleting Records — delete_id

# Delete single record
$adb->delete_id("catalog_product", 5001);

3.4 Reading Records — read_id

# Retrieve single record by ID
my @record = $adb->read_id("catalog_product", 5001);

if (@record) {
    my $id         = $record[0];  # Block 0 (ID)
    my $categories = $record[1];  # Block 1 (e.g. "5,12")
    my $brand      = $record[2];  # Block 2 (e.g. "3")
    my $authors    = $record[3];  # Block 3 (e.g. "7,9")
    my $title      = $record[4];  # Block 4
    my $price      = $record[10]; # Block 10
    print "Product: $title, Price: \$$price, Categories: $categories\n";
}

[!TIP] Best Practice: Why You Should Define Schemas (.table)
While AmberDB can operate in a schemaless mode, using schemas and index directives is strongly recommended in production and essential for maintaining speed on growing tables:

  1. Query Performance: As tables grow, queries like field_fetch and search_table rely on schema-defined indexes to execute instant direct key lookups without full disk scans.
  2. Block Layout Clarity & Living Documentation: The schema’s blocks definition provides a clear reference for your record layout (e.g., Block 1 = Category, Block 4 = Title, Block 10 = Price). It makes it easy to remember what data is stored in each block and prevents positional index confusion across developers.

(For full schema parameters and configuration rules, see Section 9: Schema Configuration)

4. Reading, Filtering, and Sorting

AmberDB provides flexible methods for listing, filtering, and sorting records.

[!CRITICAL] PAGINATION RETURN SIGNATURE & ARCHITECTURAL RATIONALE: For read_all, field_fetch, and search_table, the presence or absence of $limit governs the structure of the returned list:

⚠️ FATAL ERROR WARNING:
If you assign paginated results to a single array (my @records = $adb->read_all("catalog_product", 0, 20);), the first element $records[0] will be the integer total (e.g. 1250), not a record reference. Attempting $records[0]->[1] or $records[0][1] causes Perl to throw a fatal error: Can't use string ("1250") as an ARRAY ref while "strict refs" in use!
Rule: Whenever $limit > 0, always unpack results as my ($total, @records).

4.1 read_all — Reading All Records with Pagination

# 1. Read all records in default order (newest first - descending ID)
my @all_records = $adb->read_all("catalog_product");

# 2. Unpaginated with Extra Options (start: 0, limit: 0 returns @records / @ids directly)
# 2.1 Retrieve record IDs only (Zero deserialization, ultra memory-efficient — keys_only)
my @all_ids       = $adb->read_all("catalog_product", 0, 0, keys_only => 1);

# 2.2 Tiered Query Mode (jnktype => 'A' [Active only] | 'B' [Junk only] | 'AB' [Active + Junk])
my @active_only   = $adb->read_all("catalog_product", 0, 0, jnktype => 'A');
my @active_and_jnk= $adb->read_all("catalog_product", 0, 0, jnktype => 'AB');

# 2.3 Bypass index for direct table scan (no_index)
my @raw_records   = $adb->read_all("catalog_product", 0, 0, no_index => 1);

# 2.4 Unpaginated sorting (sort => 10 [descending] or sort => -10 [ascending])
my @all_price_asc = $adb->read_all("catalog_product", 0, 0, sort => -10); # Cheapest first
my @all_price_desc= $adb->read_all("catalog_product", 0, 0, sort => 10);  # Highest first
my @all_alpha     = $adb->read_all("catalog_product", 0, 0, sort => { blk => 4, reverse => 1 });

# 3. Paginated Queries (limit > 0 always returns ($total_count, @page))
# 3.1 First 20 records
my ($total, @page1)      = $adb->read_all("catalog_product", 0, 20);
print "Total records: $total, Retrieved on this page: " . scalar(@page1) . "\n";

# 3.2 Paginated ID list (keys_only)
my ($total, @page_ids)   = $adb->read_all("catalog_product", 0, 50, keys_only => 1);

# 3.3 Paginated and sorted
my ($total, @sorted_alpha) = $adb->read_all("catalog_product", 0, 20, sort => { blk => 4, reverse => 1 });
my ($total, @highest_price)= $adb->read_all("catalog_product", 0, 10, sort => 10);
my ($total, @lowest_price) = $adb->read_all("catalog_product", 0, 10, sort => -10);

# 3.4 Paginated and tiered (Active + Junk)
my ($total, @tiered_page)  = $adb->read_all("catalog_product", 0, 20, jnktype => 'AB');

4.2 field_fetch — Inverted Match Index (.fld) and Multi-Value Querying

Fields defined in match_block are retrieved via inverted match indexes (.fld) with O(1) average lookup time per indexed key (when querying multiple values, cost scales with the number of keys). Even if a record stores multiple comma-separated IDs (e.g. "5,12" or "7,9"), each value is indexed independently. If an index file (.fld) does not exist (unindexed tables), AmberDB seamlessly falls back to a sequential table scan (recs_scan) with identical results:

# 1. Fetch all products where Category ID (Block 1) matches "5"
my @products = $adb->field_fetch("catalog_product", 1, "5");

# 2. Fetch all products by Author ID (Block 3) "9" (Matches even if record has "7,9")
my @author_prods = $adb->field_fetch("catalog_product", 3, "9");

# 3. Paginated & sorted: Category 5 products sorted by Price (Block 10) ascending
my ($count, @sorted_prods) = $adb->field_fetch(
    "catalog_product", 
    1, "5",                             # Block 1 == "5"
    0, 12,                              # Start: 0, Limit: 12
    sort => { blk => 10, reverse => 1 } # Price ascending
);

# 4. Multi-value matching (ARRAY ref, comma-separated string, or semicolon-separated)
my @multi = $adb->field_fetch("catalog_product", 1, ["5", "8"]);
my @multi = $adb->field_fetch("catalog_product", 1, "5, 8");

# 5. Fetch scalar record IDs only (Memory-efficient pipeline)
my ($total, @id_list) = $adb->field_fetch("catalog_product", 1, "5", 0, 50, keys_only => 1);
my @all_ids           = $adb->field_fetch("catalog_product", 1, "5", keys_only => 1);

Deduplication Guarantee: Even if a record matches multiple query values simultaneously, array_nodup guarantees that each record ID appears exactly once in the result set.

2. Multi-value Match: Fetch products where Block 1 matches 5 OR 12

my @products = $adb->field_fetch(“catalog_product”, 1, “5,12”);

3. Paginated and Sorted Match: First 10 items in Category 5 sorted by price ascending

my ($total, @paged) = $adb->field_fetch( “catalog_product”, 1, “5”, 0, 10, sort => { blk => 10, reverse => 1 } );


### 4.3 `field_filter` — Multi-Criteria Faceted Filtering

Executes compound boolean queries (AND / OR) across multiple block conditions with automated bitmask intersection:

```perl
my $filter_query = {
    1  => "5",            # Category ID == 5
    2  => [ "8", "14" ],  # Brand ID IN (8, 14)
    10 => "100..500",     # Price between $100 and $500
    11 => "1",            # In Stock == 1
};

my $result = $adb->field_filter("catalog_product", $filter_query, {
    start => 0,
    limit => 20,
    sort  => { blk => 10, reverse => 1 }
});

print "Filtered Count: $result->{count}\n";
my @record_ids = @{ $result->{ids} };

4.4 search_table — Full-Text & Phonetic Keyword Search

Performs intelligent locale-aware token search across fields defined in search_block. Runs against .src inverted index files for indexed tables via direct token lookups, or performs a full table scan with identical normalization parity for unindexed tables.

# 1. Search for products matching "headphones bluetooth" (Default: AND logic)
my @results = $adb->search_table("catalog_product", "headphones bluetooth");

# 2. Paginated search with OR logic, sorted by price
my ($count, @results) = $adb->search_table(
    "catalog_product",
    "wireless headphones",
    0, 20,                                  # First 20 results
    "or",                                   # Match any keyword
    sort => { blk => 10, reverse => 1 }     # Sort by price ascending
);

# 3. Retrieve only matching record IDs (keys_only)
my ($count, @id_list) = $adb->search_table("catalog_product", "sony", 0, 50, keys_only => 1);
my @all_ids           = $adb->search_table("catalog_product", "sony", keys_only => 1);

Key Highlights of AmberDB Search Normalization:

4.5 read_list — Reading Specific IDs in Specified Sequence

read_list is AmberDB’s high-throughput batch record resolution engine. It plays an essential role both in the engine’s internal query pipeline and in developer application code:

1. Internal Engine Pipeline:

All high-level listing and querying methods in AmberDB (read_all, field_fetch, search_table, field_filter, etc.) operate in two decoupled stages:

  1. Index Filtering Stage: The query method first reads lightweight record keys (@ids) from inverted index files (.inx, .fld, .src, .srt), evaluating Boolean logic (AND/OR), sorting, and pagination slicing (recs_cutting).
  2. Batch Document Resolution Stage: Once the final matched ID list is finalized, it is forwarded in a single call to read_list. read_list opens the data table in a single batch session (or leverages the RAM-Disk cache) to deserialize all requested records simultaneously, returning them in the exact positional order requested.

2. Developer API Usage & Relational Traversal (SQL JOIN Alternative):

Developers can use read_list directly to retrieve full document records for arbitrary collections of IDs efficiently in a single operation.

Example Scenario: Fetching Full Profiles of Customers with Active Orders

# 1. Retrieve all active order records
my @orders = $adb->read_all("order_active");

# 2. Assume Block 2 of each order record ($order[N]->[2]) holds the Customer ID.
# Extract unique Customer IDs using map:
my %customer_ids = map { $_->[2] => 1 } @orders;

# 3. Fetch full profile records for all matching customers in a single batch call:
my @customer_records = $adb->read_list("customers", [ keys %customer_ids ]);

foreach my $customer (@customer_records) {
    my $c_id      = $customer->[0]; # Customer ID
    my $c_name    = $customer->[1]; # Full Name
    my $c_email   = $customer->[2]; # Email
    my $c_address = $customer->[3]; # Delivery Address (Shipping label / dispatch list)
    print "Shipping Label -> ID: $c_id | Name: $c_name | Email: $c_email | Address: $c_address\n";
}

[!TIP] read_list accepts an array reference (\@ids) or a flat array. It guarantees that the returned records preserve the exact sequential order of the input ID list.

4.6 Existence Check Functions

Quickly check whether a record or table exists without pulling full data into memory:

# 1. Single Record Existence (O(1) direct key check)
if ($adb->exist_id("catalog_product", 5001)) {
    print "Product 5001 exists in database.\n";
}

# 2. Bulk Existence Check
my $presence_map = $adb->exist_list("catalog_product", 5001, 5002, 9999);
# Returns: { 5001 => 1, 5002 => 1, 9999 => 0 }

# 3. Physical Table / File Existence
if ($adb->exist_table("catalog_product")) {
    print "catalog_product.db exists on disk.\n";
}

# Check specific file extension (e.g. .slg slug map)
if ($adb->exist_table("catalog_product", "slg")) {
    print "Slug index file exists.\n";
}

4.7 Positional and Special Reads — read_firstid, read_lastid, read_randid, and read_count

# 1. Read First Record by Numeric Key Order
my @first_item = $adb->read_firstid("catalog_product");

# 2. Read Last (Latest Added) Record
my @latest_item = $adb->read_lastid("catalog_product");

# 3. Read Random Record (Daily deal / Random featured product)
my @random_item = $adb->read_randid("catalog_product");
print "Featured Deal: $random_item[4] (\$$random_item[10])\n";

# 4. Read View / Hit Counter from .cnt File
my $views = $adb->read_count("catalog_product", 5001);
print "Product 5001 viewed $views times.\n";

5. Simple Mode and Direct Schemaless Access (Simple Mode)

In AmberDB, Simple Mode (simple => 1) represents the entirely schemaless, lightweight, direct flat-file NoSQL operational mode where no .table or .dbase schema files and no secondary binary indexes (.inx, .src, .fld, .fac, .srt, .slg, .aut, .del) are generated or maintained.

In Simple Mode, records can store rich, nested data structures directly, including array and hash references (ARRAY/HASH). The index generation and maintenance overhead is completely eliminated; single-key read and write operations (read_id, insert_id) execute at maximum hardware speed ($O(1)$).

5.1 Initializing and Activating Simple Mode

Simple Mode can be activated in four distinct ways:

  1. Constructor Initialization via cfg: 
    my $adb = AmberDB->new(
    path => { dbase_dir => "/var/data/sessions" },
    cfg  => { simple    => 1 },
);
  2. Quick Shortcut Helper via AmberDB::Tools (db_simple): 
    use AmberDB::Tools;
my $tools = AmberDB::Tools->new();
my $adb   = $tools->db_simple("/var/data/sessions");
  3. Dynamic Runtime Switch via config: 
    $adb->config( simple => 1 );
  4. Automatic Simple Mode Trigger via Custom Extensions (db_ext):
    AmberDB defaults to .db. If db_ext is configured with any extension other than "db" (e.g. "dat", "cache", "session"), the engine automatically switches into Simple Mode: 
    my $adb = AmberDB->new(
    path => { dbase_dir => "/var/data/cache" },
    cfg  => { db_ext    => "dat" },  # Automatically activates simple => 1
);

Directory Layout Note: In standard mode, tables reside under $dbase_dir/tables/. In Simple Mode, the engine creates and reads database files directly inside the root of dbase_dir ($dbase_dir/<table_name>.<ext>). To open existing standard-mode tables in simple mode, set dbase_dir directly to dbstore/tables.

5.2 Flexible & Arbitrary Record IDs (No 8-Byte Limit)

The standard mode 8-byte limit and strict ASCII/numeric format constraints are relaxed in Simple Mode (id_check accepts arbitrary scalar keys and applies safe key sanitization):

$adb->insert_id( 'sessions', 'user@example.com', 'Active', 'Chrome', time() );
my @sess = $adb->read_id( 'sessions', 'user@example.com' );

5.3 Data Operations (CRUD & Bulk)

All standard CRUD and bulk methods operate seamlessly in Simple Mode:

# Single Insert, Read, Modify, Delete
$adb->insert_id( 'orders', 'order_101', 'Pending', '150.00' );
my @order = $adb->read_id( 'orders', 'order_101' );
$adb->modify_id( 'orders', 'order_101', 'Completed', '175.50' );
$adb->delete_id( 'orders', 'order_101' );
my $exists = $adb->exist_id( 'orders', 'order_101' );

# Bulk Operations (Bulk CRUD)
my $ins_status = $adb->insert_list( 'orders', [ 'o_1', 'A', 50 ], [ 'o_2', 'B', 75 ] );
my $mod_status = $adb->modify_list( 'orders', [ 'o_1', 'A+', 55 ] );
my $del_status = $adb->delete_list( 'orders', 'o_1', 'o_2' );

5.4 Unindexed Direct Queries & Filtering

Since secondary index files are omitted, queries stream sequentially across the raw database file (recs_scan):

  1. Table Scan and Pagination (read_all): 
    # Paged scan (start => 0, limit => 10)
my ( $total_count, @records ) = $adb->read_all( 'items', 0, 10 );
   
# Retrieve keys only
my @keys = $adb->read_all( 'items', keys_only => 1 );
   
# In-memory sorting (Block 3 ASC: -3, DESC: 3)
my @sorted = $adb->read_all( 'items', sort => -3, keys_only => 1 );
  2. Field Value Fetching (field_fetch): 
    # Block 2: Category = 'Apparel'
my @apparel = $adb->field_fetch( 'catalog', 2, 'Apparel' );
   
# Multi-value matching (Block 3: Color in ['Blue', 'Black'])
my ( $cnt, @results ) = $adb->field_fetch( 'catalog', 3, [ 'Blue', 'Black' ], 0, 20, sort => -4 );
  3. Full-Text Word Search (search_table): 
    # Collation-aware word search (AND logic)
my @articles = $adb->search_table( 'articles', 'market economy' );
   
# Combined search with field filter (Block 2: Category = 'Finance')
my ( $cnt, @filtered ) = $adb->search_table( 'articles', 'rates', 0, 10, filter => [ 2, 'Finance' ] );

5.5 ACID Transactions

In Simple Mode, transact_start, transact_commit, and transact_rollback provide full ACID transaction safety. When a rollback is triggered, raw modifications in the .db file are restored:

$adb->transact_start();
eval {
    $adb->insert_id( 'sessions', 'token_123', 'TempData', time() );
    die "Critical error" if $failed;
    $adb->transact_end();
};
if ($@) {
    $adb->transact_rollback(); # token_123 is cleanly reverted from the .db file
}

5.6 Continuous Daily Backup Logs (recs_back)

Because text backup is schema-independent, daily audit and continuous recovery streaming (recs_back) is fully active in Simple Mode.

In accordance with Simple Mode’s flat directory structure, no separate backup/ or YYYY/ subfolder is created. Every insert_id (add), modify_id (edit), and delete_id (del) operation is logged directly to $dbase_dir/YYYY-MM-DD.csv in the same directory alongside database tables:

2026-08-31 14:30:00    admin    add     sessions    sess_token_99999    Active\x1f192.168.1.50
2026-08-31 14:31:15    admin    edit    sessions    sess_token_99999    Closed\x1f192.168.1.50
2026-08-31 14:32:00    admin    del     sessions    sess_token_99999

5.7 RAM-Disk Architecture & Caching in Simple Mode

In standard mode, AmberDB manages RAM-disk staging via schema use_cache => 2 rules.

In Simple Mode, RAM-disk utilization is direct and flexible:
Since Simple Mode requires no schema files, creating a high-performance in-memory cache or session store simply involves binding a second AmberDB instance directly to the RAM-disk / tmpfs mount:

# 1. Persistent disk instance (For durable storage)
my $db_disk = AmberDB->new(
    path => { dbase_dir => "/var/data/app/dbstore/tables" },
    cfg  => { simple => 1 },
);

# 2. RAM-Disk instance (Zero-latency in-memory cache/session store)
# (Linux: /dev/shm or tmpfs, Windows: ImDisk / RamDisk volume)
my $db_ramdisk = AmberDB->new(
    path => { dbase_dir => "/dev/shm/amber_cache" },
    cfg  => { simple => 1, no_backup => 1 }, # Disable backup for pure transient cache
);

# In-memory reads and writes at nanosecond speed:
$db_ramdisk->insert_id( "sessions", $session_token, $user_id, time() );
my @sess = $db_ramdisk->read_id( "sessions", $session_token );

Benefits of this dual-instance design:

5.8 Feature Comparison: Standard vs. Simple Mode

Feature / Subsystem Standard Mode (simple => 0) Simple Mode (simple => 1)
Schema Files (.table, .dbase) Required & Enforced None / Schemaless
Arbitrary & Long Record IDs 8-Byte / Strict ASCII Limits Completely Unrestricted
Direct CRUD (insert_id, read_id) $O(1)$ $O(1)$ (Max Throughput)
Bulk Operations (insert_list, etc.) Supported Supported
Table Scan (read_all) Binary .inx or Direct Direct Streaming Scan
Pagination (limit) & keys_only Supported Supported
In-Memory Sorting (sort => 2) Supported Supported
Field Matching (field_fetch) Indexed .fld $O(1)$ Sequential Streaming Scan
Word Search (search_table) Inverted Index .src Collation Streaming Scan
ACID Transactions (transact_*) Supported (Index Undo) Supported (Raw Undo)
Continuous Daily Backup (recs_back) Supported (backup/YYYY/) Supported (Same Directory YYYY-MM-DD.csv)
Secondary Indexes (.inx, .fld, .src, .srt, .fac) Generated & Maintained Disabled (Zero Index Cost)
URL Slug Mapping (.slg) Auto Generated Disabled
Audit Logs (.aut) & Archive (.del) Schema-Driven Disabled
Directory Hierarchy tables/, schema/, backup/, etc. Flat Single Directory ($dbase_dir/<table_name>.db)
Secondary Indexes (.inx, .fld, .src, .srt, .fac) Generated & Maintained Disabled (Zero Index Cost)
URL Slug Mapping (.slg) Auto Generated Disabled
Audit Logs (.aut) & Archive (.del) Schema-Driven Disabled
Directory Hierarchy tables/, schema/, backup/, etc. Flat Single Directory ($dbase_dir/<table_name>.db)

6. Indexing and Search Engine

AmberDB maintains structured binary index files based on the schema configuration.

6.1 Index Types

Extension Index Type Description
.inx Record Index Packed binary array of all active IDs, total count, and highest ID.
.fld Match Index Block-level key-to-IDs inverted index (field_fetch).
.str Field Dictionary Bidirectional string-to-numeric ID dictionary companion for .fld (_${blk}.str).
.src Full-Text Index Word-level token inverted index (search_table).
.srt Sort Index Pre-sorted binary array of record IDs for sort_block definitions.
.fac Facet Index Fast forward index for faceted filter navigation.
.slg URL Slug Index Bidirectional map: _0.slg (ID → Slug) and _1.slg (Slug → ID).

6.2 Unified 8-Byte Binary Packing Standard

AmberDB achieves high throughput and compact disk storage through uniform 8-byte binary packing:

This binary layout enables zero-copy slicing for pagination (LIMIT/OFFSET) directly through raw byte offsets ($O(1)$ substr slicing) without decoding full record buffers into memory.

6.3 Inverted Match Index (.fld) and Bidirectional Dictionary (.str)

For fields declared under match_block, AmberDB indexes data across two complementary tiers:

  1. Packed Binary Inverted Match Index (.fld):
    Maintains a dedicated <table_name>_<blk>.fld file per block. Keys map directly to 8-byte packed binary arrays (Q* / a8*) containing matching record IDs. Queries via field_fetch perform direct $O(1)$ key lookups into this file.

  2. Bidirectional String-to-ID Dictionary (.str):
    For non-relational free-text attributes (Category Name, Brand Name, Author, Status Tags), the engine automatically manages a companion <table_name>_<blk>.str dictionary:

    • Forward Lookup (s:<term> $\rightarrow$ $nid): Assigns an incremental numeric token ID to each unique textual string.
    • Reverse Lookup (n:$nid $\rightarrow$ <term>): Enables $O(1)$ reverse label translation from numeric IDs back to human-readable text.
    • Transparent Resolution: When calling field_fetch or field_filter, developers can pass either the canonical numeric ID (12) or the textual label ("Sony"). The engine automatically resolves text terms via .str and retrieves the matching records from .fld.

6.4 Sorting Mechanism & Developer Guide

AmberDB provides high-performance, pre-indexed sorting across specific table blocks.

6.4.1 Schema Configuration (sort_block)

Define sortable blocks in your .table schema file. Specify a simple block index (4), or declare explicit types (type) for numeric and date fields:

# dbstore/schema/catalog_product.table
{
    id_type    => 'num',
    sort_block => [
        4,                             # Block 4: Title (String sorting)
        { blk => 10, type => 'num' },  # Block 10: Price (Numeric sorting)
        { blk => 12, type => 'date' }, # Block 12: Timestamp sorting (YYYYMMDDHHMMSS)
    ],
}

6.4.2 Using Sort in Query Methods

Pass the sort option to read_all, field_fetch, or search_table to retrieve sorted datasets immediately:

# 1. Default Direction: Descending / Highest First (DESC: 99->0, Z->A)
my @products = $adb->read_all("catalog_product", sort => 10);
my @products = $adb->read_all("catalog_product", sort => { blk => 10 });

# 2. Reverse Direction: Ascending / Lowest First (ASC: 0->99, A->Z)
my @products = $adb->read_all("catalog_product", sort => -10);
my @products = $adb->read_all("catalog_product", sort => { blk => 10, reverse => 1 });

# 3. Primary Key (ID) Ascending Order:
my @products = $adb->read_all("catalog_product", sort => { reverse => 1 }); # 1..N oldest first

# 4. Sorting with field_fetch and search_table:
my @cat_items       = $adb->field_fetch("catalog_product", 1, "electronics", sort => { blk => 10, reverse => 1 });
my ($count, @search) = $adb->search_table("catalog_product", "headphone", 0, 20, sort => -10);

7. Transaction Safety, ACID Guarantees, and Crash Recovery (Transactions)

AmberDB::Transact provides full ACID-compliant transactions and Strict Two-Phase Locking (Strict 2PL) concurrency control for multi-table updates (e.g., creating an order, updating inventory, and charging accounts).

7.1 Transactional Integrity & The Single Transaction Spine

In modern e-commerce and enterprise workflows, a single high-level user action (such as “Complete Checkout”) triggers an interdependent semantic operation chain spanning multiple tables and sub-systems:

Checkout Operation Chain:
 ├─ Order Confirmation (creating entry in orders table)
 ├─ Cart Cleared (purging items from cart table)
 ├─ Customer Account (balance deduction or card charge record)
 ├─ Company Account (revenue entry in general ledger)
 ├─ Stock Inventory (deducting counts in catalog_product table)
 └─ Supplier Dispatch (writing work item to supplier_queue table)

These operations are semantically coupled and mutually dependent. If one operation fails while the others persist, the database falls into an inconsistent state:

the system state becomes corrupted. To eliminate these anomalies, the entire sequence must be unified within a single transaction spine (transact_start $\rightarrow$ transact_end). If any step encounters an error or if the process crashes, AmberDB evaluates the .txn undo log in reverse (LIFO) order, completely rolling back all modified tables and secondary indexes to their pristine pre-transaction state.

7.2 ACID Guarantees in AmberDB

AmberDB guarantees the four classical ACID properties through embedded flat-file database mechanics:

ACID Property Implementation Mechanism & Guarantees
Atomicity Disk-Backed Undo-Journaling: When transact_start() is called, a microsecond-stamped .txn journal is created. Every insert_id, modify_id, and delete_id call appends reverse undo instructions. If a critical base error occurs or transact_rollback() is triggered, changes across base records (.db), soft-delete archives (.del), user audit logs (.aut), and all secondary indexes (.inx, .src, .fld, .fac, .srt, .slg, .jinx, .jsrc, .jfld) are completely reverted in reverse LIFO order.
Consistency Schema, Index, and State Integrity: Inbound records are validated against schema field rules, data types, and byte limits. Primary keys (autoid), inverted word indexes, columnar facets, and URL slugs are synchronized in real time. Upon rollback, both in-memory caches (cache_delete) and secondary indexes revert to their clean pre-transaction state, preventing corrupted intermediate states.
Isolation Strict Two-Phase Locking (Strict 2PL): Every record modified within an active transaction acquires an exclusive OS-level lock (flock LOCK_EX). Locks are held throughout the entire transaction duration, preventing concurrent workers from modifying the locked records. Locks are released simultaneously only upon commit or rollback, providing serializable isolation.
Durability Synchronous Journaling & Crash Recovery (transact_recover): All journal writes invoke $fh->flush. When configured with cfg => { txn_sync => 1 }, AmberDB triggers OS/kernel fsync ($fh->sync) and Berkeley DB cache flushing (DB_File->sync). If a process or server crashes mid-transaction, orphaned .txn files are detected via non-blocking flock checks and rolled back automatically.

Architectural Note: Batch ETL Imports vs. Business Transactions
Methods such as insert_list, modify_list, and delete_list are specialized for high-throughput batch imports (e.g., ingesting large XML/JSON product catalogs). Since list records are typically independent entities without cross-dependencies, discarding thousands of valid records due to a few malformed entries in such bulk ingests is undesirable. For cyclical business logic where interdependent operations must succeed or fail as a single atomic unit (orders, inventory, billing), use single-record CRUD methods within a transact_start / transact_end block. If a list ingestion strictly requires full transactional rollback, place the dataset inside a loop executing single-record operations (insert_id, modify_id, delete_id) within a transaction block so the entire batch is fully transacted.

7.3 Transaction Workflow

  1. transact_start(): Opens a microsecond-stamped undo journal (.txn) in $dbase_dir/txn/ and recovers any orphaned transactions left by dead processes (transact_recover).
  2. CRUD Operations: insert_id, modify_id, delete_id write updates to the base .db file, acquire record write locks (flock), and record reverse undo entries in the .txn journal.
  3. transact_end(): Finalizes the transaction.
    • If clean: Deletes journal, releases held locks, and commits changes (status => "commit").
    • If base errors occurred: Evaluates journal in reverse (LIFO) order, restoring base records and index states to their pre-transaction snapshot, releases locks (status => "rollback").
  4. transact_rollback(): Manually triggers immediate rollback based on business logic.

7.4 Practical Example: Checkout & Inventory Transaction

# 1. Start Transaction
$adb->transact_start();

my $product_id = 42;
my $quantity   = 2;
my $user_id    = 1001;

# Read product and check inventory
my @product = $adb->read_id("catalog_product", $product_id);
my $current_stock = $product[8]; # Block 8 = Stock count

if ($current_stock < $quantity) {
    # Insufficient stock: report transaction error (transact_end will trigger automatic rollback)
    $adb->transact_error("catalog_product", "Insufficient stock ($current_stock < $quantity)");
} else {
    # Deduct stock and update product (@product[0] contains $product_id)
    $product[8] -= $quantity;
    $adb->modify_id("catalog_product", @product);

    # Create order record
    my @order = ( $user_id, $product_id, $quantity, time(), "confirmed" );
    my $order_id = $adb->insert_id("orders", undef, @order);
}

# Finalize transaction (commits if clean, automatically rolls back on error)
my $res = $adb->transact_end();

if ($res->{status} eq "commit") {
    print "Order placed successfully and stock deducted!\n";
} else {
    warn "Transaction aborted! All changes were automatically rolled back.\n";
}

7.5 Durability and Crash Recovery

7.6 Core Architectural Philosophy: Authoritative Data vs. Rebuildable Indexes

AmberDB’s storage and transaction architecture is organized around a strict hierarchy of data authority:

  1. Authoritative Master Files (Non-Reconstructible Source of Truth):
    • .db (Master Document Data): Primary storage for all active records and documents.
    • .del (Soft-Deleted Archive): Preserves deleted records under keep_deleted. Once moved here, deleted data cannot be reconstructed from .db.
    • .aut (User Audit Trail): Chronological, time-series history of who created, edited, or deleted records (log_owner). This historical data cannot be generated from any other source.
  2. Derived & Rebuildable Indexes (Disposable Secondary Projections):
    • .inx (Record Index), .fld (Match), .src (Full-Text), .srt (Sort), .fac (Facet), .slg (URL Slug): All these index files are deterministic projections derived directly from .db.
    • If any secondary index is corrupted, deleted, or incomplete, running AmberDB::Tools->set_index($table) reconstructs all indexes from scratch within seconds with zero data loss.

Rationale Behind Transaction Design: AmberDB::Transact was deliberately engineered around this principle. A failure writing to the authoritative .db file (is_index == 0) triggers an immediate automatic rollback. However, if the master document is safely committed to .db and an index update encounters a disk error (is_index == 1), valid business data is never discarded; the transaction commits, and indexes can simply be repaired using AmberDB::Tools.

7.7 Exempting Auxiliary Tables from Failure Cascades (no_transact)

In multi-table business operations (e.g. creating an order, updating inventory, and charging accounts), some tables represent core transactional entities (orders, payments, inventory), while others serve as auxiliary or secondary records (customer order summaries, product view counters, notification queues). An unexpected failure writing to an auxiliary table should not abort or roll back a successfully charged order.

AmberDB allows declaring tables with no_transact => 1 (either in schema .table or dynamically at runtime) to exempt them from transaction abort cascades:

  1. Static Schema Definition (.table file): 
    # order_customer_summary.table
{
    name        => "Customer Order Summary",
    no_transact => 1,   # Failures here do NOT abort the main transaction
    schema      => [qw(user_id order_id amount created_at)],
}
  2. Dynamic Runtime Configuration (table_attr): 
    # Temporarily exempt an auxiliary table during a specific workflow:
$adb->table_attr("order_customer_summary", no_transact => 1);

How It Works:

7.8 Multi-Process Concurrency, Lock Isolation, and Stress Verification

AmberDB is engineered for high-concurrency production deployments (Apache, Plack/PSGI, Starman, FastCGI, Starlet) and background worker pools (cron jobs, async queues) where dozens of independent processes simultaneously read and write to the same database tables and secondary indexes.

Operating System-Level Lock & Platform Isolation:

  1. Linux / POSIX Environments: Leveraging POSIX fork() and kernel-level flock(LOCK_EX) / flock(LOCK_SH) locks, every worker process operates within an isolated memory address space. Strict Two-Phase Locking (Strict 2PL) guarantees serializable transaction isolation across processes.
  2. Windows / MSYS2 Environments: On Windows NT architectures, full OS-level lock integrity and file descriptor isolation are maintained across independent worker processes (perl.exe).
  3. Record-Level Locking (flock_open / flock_close): For critical concurrent updates on individual records (such as high-demand stock decrements or shared counters), $adb->flock_open($table, "write", $id) eliminates race conditions and lost updates with 100% precision.

Concurrency & Stress Test Suite (xt/amberdb_concurrency_stress.t):

The database engine’s resilience under extreme parallel load is verified by the author/release stress test suite:

# Run multi-process concurrency stress tests directly:
perl -Ilib xt/amberdb_concurrency_stress.t

This test suite validates 5 mission-critical concurrency scenarios:

8. High-Throughput Batch Operations (Batch ETL & Ingestion)

AmberDB provides a dedicated 2-Phase Batch Pipeline for ingesting and updating large volumes of records (ETL from CSV, JSON, XML, or REST APIs) at maximum throughput.

8.1 Why Use insert_list Instead of insert_id in a Loop?

Executing insert_id in a loop forces the operating system to perform $N$ independent file opens (open/tie), lock acquisitions (flock), auto-increment sequence mutations, and secondary index writes (.inx, .src, .fld, .fac, .srt). For $N$ records, this incurs $O(N \times K)$ file I/O operations and process context switches.

insert_list splits the ingestion workflow into 2 unified phases, reducing I/O complexity to $O(K)$:

  1. Phase 1 (Single I/O Master Table Write): The .db Berkeley DB file is opened exactly once (table_write). Auto-increment IDs are allocated contiguously (table_autoid), field formatters and schema rules are evaluated, and all records are flushed into the hash table in one single stream (recs_put).
  2. Phase 2 (Batched Secondary Index Merge): Each secondary index file (.inx, .src, .fld, .fac, .srt, and junk tier) is opened exactly once and the entire batch is compiled into binary bitsets and B-tree branches via unified merges (records_add, search_add, match_add, facet_add, sort_add).

[!TIP] On a batch of 10,000 records, insert_list finishes 50x to 100x faster than a standard insert_id loop.

8.2 Batch Insert (insert_list)

# Array of record column tuples. 
# Pass 0 or undef for ID to automatically allocate 64-bit auto-increment IDs.
my @new_products = (
    [ 0, "5",    "3", "Wireless Headphones", "149.90", "2026-08-28", "1" ],
    [ 0, "5,12", "8", "Mechanical Keyboard", "299.00", "2026-08-28", "1" ],
    [ 0, "12",   "3", "Gaming Mouse",        "89.50",  "2026-08-28", "1" ],
    # ... hundreds or thousands of records ...
);

my $status = $adb->insert_list("catalog_product", @new_products);
# Returns hashref of created IDs: { 101 => 1, 102 => 1, 103 => 1, ... }

8.3 Bulk Modify (modify_list)

my @updates = (
    [ 101, "5",    "3", "Wireless Headphones Pro", "179.90", "2026-08-28", "1" ],
    [ 102, "5,12", "8", "Mechanical Keyboard RGB",  "329.00", "2026-08-28", "1" ],
);

my $status = $adb->modify_list("catalog_product", @updates);

8.4 Bulk Delete (delete_list)

# Target IDs can be passed as a flat list or array reference
my $status = $adb->delete_list("catalog_product", 101, 102, 103);
# or:
# $adb->delete_list("catalog_product", [101, 102, 103]);

8.5 Chunking Strategy for Large Datasets (ETL Ingestion)

For massive imports (e.g. 50,000+ records), chunking records into batches of 500 to 1,000 optimizes memory allocation and balances disk cache flushing:

my $chunk_size = 1000;
for (my $i = 0; $i < @huge_dataset; $i += $chunk_size) {
    my $end = $i + $chunk_size - 1;
    $end = $#huge_dataset if $end > $#huge_dataset;
    my @chunk = @huge_dataset[$i .. $end];
    $adb->insert_list("catalog_product", @chunk);
}

9. Schema Configuration (.table & In-Memory)

AmberDB is a schema-driven database engine. Table schemas define primary key constraints, field data types, multi-dimensional indexes, automatic URL slug generation, facet filters, lifecycle junk rules, data validation constraints, and variable repeating nested child records that eliminate SQL JOIN bottlenecks.

9.1 Database and Table Directory Layout

AmberDB stores tables, indexes, and schema definitions in dedicated physical directories under the configured dbstore root:

Directory Purpose
dbstore/tables/ Base data (.db) and binary indexes (.inx, .fld, .src, .fac, .srt, .slg)
dbstore/schema/ Schema files (.table) and group configs (.dbase)
dbstore/conf/ Plain-text .conf configuration and property files
dbstore/backup/ Daily CSV audit backups (dbgun/YYYYMMDD/)
dbstore/cache/ Unified Shared RAM-Disk (ImDisk/tmpfs) Root:
dbstore/cache/tables/ Mirrored hot .db and .inx tables in RAM for use_cache => 1 & 2
dbstore/cache/conf/ Compiled high-speed config cache (*.pl hash references)
dbstore/cache/schema/ Cached / pre-compiled table schemas in RAM (*.table, *.dbase)
dbstore/cache/lock/ Process and table-level flock lock files in RAM (*.lock)
dbstore/cache/pids/ Process lock files and login error state logs (*.pid, *.error)

[!IMPORTANT] Version 5.21.0 Migration Notice: The only manual action required when upgrading existing projects is to rename your database directory’s dbstore/scheme/ folder to dbstore/schema/. All programmatic path resolutions and API calls are automatically handled by the engine.

9.2 Schema Role & Flexibility: Optional vs. Full Definition

Schema design in AmberDB is modular, tiered, and highly flexible:

# dbstore/schema/catalog_product.table
{
    name         => "Product Catalog",
    id_type      => "num",                  # "num" (64-bit uint) or "ascii" (max 8 bytes)
    record_index => 1,                      # Enable .inx primary record index & auto-increment counter
    match_block  => [ 1, 2, 3, 11 ],        # .fld Exact field match indexes (Category, Brand, Author, Status)
    search_block => [ 4, 5, 7, 9 ],         # .src Full-text search fields (Title, Subtitle, Description, Barcode)
    sort_block   => [ 4, { blk => 10, type => 'num' } ], # .srt Pre-sorted binary ID buffers
    keep_deleted => 1,                      # Preserve soft-deleted record timestamps in .del
    log_owner    => 1,                      # Write operator audit trails to .aut log
}

9.3 Schema Definition & Retrieval Methods (table_info & table_attr)

  1. Disk-Based Schemas (Recommended):
    Placed in dbstore/schema/<table_name>.table. AmberDB automatically parses and caches them on first access.

  2. In-Memory Dynamic Schemas:
    Programmatically assigned at runtime via $adb->table_attr("table_name", { ... }).

  3. Retrieving Active Schema (table_info):
    To inspect the parsed configuration hash reference for any table, call $adb->table_info($table_name):

    my $schema = $adb->table_info("catalog_product");
print "Table Name: $schema->{name}\n";
print "Search Blocks: " . join(", ", @{ $schema->{search_block} || [] }) . "\n";

[!IMPORTANT] Schema Files (.table and .dbase) Are Native Perl Code (Hash References)
In AmberDB, .table and .dbase files are not static JSON or YAML documents; they are native Perl hash references ({ ... }) dynamically evaluated at runtime via Perl’s built-in do statement.

9.4 Table Naming Conventions

9.5 Example Schema (catalog_product.table)

# dbstore/schema/catalog_product.table
{
    name         => "Product Catalog",
    id_type      => "num",
    record_index => 1,
    match_block  => [ 1, 2, 3 ],
    search_block => [ 4, 5 ],
}

9.6 Schema Configuration Parameters Reference (Table Level)

The following reference table details all top-level parameters supported in .table schema definitions, along with default values and legacy alias equivalents:

Parameter Type Default Legacy / Alias Description
name string "Table" — Human-readable table title.
id_type string "num" — Primary key format: "num" (64-bit unsigned int) or "ascii" (max 8-byte alphanumeric string).
record_index 0 / 1 0 readall When 1, enables the .inx primary binary index, table_count, table_lastid, and auto-increment.
search_block ARRAY [] — Block numbers indexed in .src for full-text inverted search.
match_block ARRAY [] fields Block numbers indexed in .fld for exact field-to-ID matching and relational lookup.
sort_block ARRAY [] — Pre-computed .srt binary sort indexes ([ 4, { blk => 10, type => 'num' } ]).
facet_block ARRAY [] filter_block Block numbers indexed in .fac for columnar faceted category navigation.
slug_block ARRAY [] rwlink Block numbers combined for automated bidirectional .slg URL slug generation (e.g. [2, 4]).
use_facet 0 / 1 0 — Enables the facet counting engine and field_fltkeys / facet_menu on the table.
facet_rules ARRAY [] — Scoping rules for facet counting (e.g., displaying only in-stock items in filter menus).
use_junk 0 / 1 0 — Enables dual-tier indexing by segregating inactive/out-of-stock records to Cold Tier B.
junk_rules ARRAY [] — Business rules determining automatic routing of records between active and junk tiers.
use_cache 0 / 1 / 2 0 usecache 0: Disabled, 1: Soft (.inx metadata), 2: Hard (Full shared RAM-Disk mirror).
cache_ttl integer 3600 — Table-specific RAM cache time-to-live in seconds.
keep_deleted 0 / 1 0 nodelete Preserves deleted records in .del soft-delete archive instead of permanent deletion.
log_owner 0 / 1 0 authority Records user modification audit trails in .aut files.
use_alias 0 / 1 0 uselnk Enables .lnk alias routing table for merged records or legacy URL redirections.
use_counter 0 / 1 0 usecnt Enables automated hit/view read counters in .cnt files.
parent_table string "" — Parent table name for vertical partitioning (child table shares the same primary ID).
force 0 / 1 0 — When 1, insert_id overwrites existing records rather than failing (Replace mode).
min_char integer 2 minchar Minimum word length for full-text search indexing (1, 2, or 3).
stop_word string "" nextkey Stop-words excluded from full-text search indexing (e.g., "the and for with").
repeat_ids integer undef — Target block number where extracted child item IDs are consolidated.
repeat_start integer undef — Starting block index for dynamic repeating child rows (order items, cart lines).
view_block ARRAY [] — Priority block numbers displayed in UI / CMS listing views.
use_menu 0 / 1 1 — Controls display of the table in admin panel navigation menus.
no_transact 0 / 1 0 — Exempts table from transactional rollback error propagation.
no_backup 0 / 1 0 — Disables daily CSV user audit logging for this table.

9.7 Block (Field) Definitions, 8 Core Field Types, UI Inputs, and Validation Reference

Each block definition inside the blocks array supports the following attributes:

9.7.1 Core Block Attributes

Attribute Type Description Example
id string Programmatic field identifier id => "email"
name string Display label for UI forms and table headers name => "Email Address"
type string Data storage, type validation, and indexing type type => "text"
input string HTML/UI Form input component type input => "select"
valid string Automated data validation rule valid => "not_null;email"
option string Enumerated choice options (value:label pairs) option => "1:Active,0:Inactive"
rdbm string / HASH Foreign table lookup mapping (foreign_table;display_block) rdbm => "catalog_category;2"
extend HASH 1:1 vertical table extension extend => { table => "catalog_price", join => "id" }

9.7.2 Supported 8 Core Field Types (type)

AmberDB uses 8 unified core storage types across serialization (db_encode/db_decode), indexing, and sorting layers:

Field Type (type) Description enc_validate (Write Phase) dec_validate (Read Phase) Indexing & Sorting Behavior
auto_id Auto-increment ID (Block 0) Primary key format validation ID scalar return Primary key index (.inx)
text Standard UTF-8 Text UTF-8 string validation String scalar ($val // '') Inverted index (.src), dictionary (.str)
num / number Numeric (Integer / Float / Boolean) Numeric validation (^[+-]?[0-9]+(?:\.[0-9]+)?$), defaults empty to 0 Numeric scalar cast (0 + $val) Numerical sorting (<=>) in .srt, .fld filters
ascii ASCII-Only Text ASCII normalization via to_ascii Clean ASCII text URL slug map (.slg), ASCII .srt sorting
date Date and Time Assigns system date if auto_date is active Date string Chronological sort in .srt via str2dateid
array / repeat List / Repeating Rows ARRAY ref or [split /,/] Perl ARRAY ref ([]) Multi-value matching (field_fetch)
hash Dictionary / Object (HASH ref) HASH ref validation Perl HASH ref ({}) Schemaless nested key-value store
binary Binary Payload / Base64 Raw bytes or Base64 string Raw binary scalar Direct flat file storage

[!NOTE] Numeric and Boolean Management: The num (or number) type handles positive (150, +25), negative (-50, -12.75), floating-point values, and 0 / 1 boolean flags. Unchecked HTML checkboxes or empty numerical inputs are automatically normalized to 0 by enc_validate and dec_validate.

9.7.3 UI Form Input Components (input)

Determines how the field is rendered in UI forms and administration panels:

Component (input) UI Element Description
text Text Input Standard single-line text field <input type="text">.
textarea Textarea Multi-line plain text box <textarea>.
summernote Summernote Rich WYSIWYG HTML visual editor for articles/descriptions.
select Dropdown Select Single-selection dropdown list <select>.
checkbox Checkbox Multi-selection checkboxes <input type="checkbox"> (Use type => "num" for Boolean).
radio Radio Buttons Single-selection radio options <input type="radio">.
file File Upload Attachment or image file uploader <input type="file">.
hidden Hidden Field Hidden form element <input type="hidden"> (for primary IDs).
email Email Input HTML5 email input field <input type="email">.
ascii ASCII Field User/code input box constrained to ASCII charset.
number Number Input Numeric stepper <input type="number">.
date Date Picker Interactive date calendar selector <input type="date">.
password Password Field Obscured security input <input type="password">.
repeat / repeats Repeater Table Dynamic sub-row table input with add/remove row buttons (Order items, invoice lines).
search_block Search Box Search-assisted dynamic filter input.
selectbyfind SelectByFind Foreign relation selector populated via dynamic search.
selectbylist SelectByList Multi-item picker component from list.

9.7.4 Repeating Child Row Blocks (repeat_start and repeat_ids)

AmberDB natively supports dynamic repeating child rows (e.g. order line items, invoice product rows) horizontally across the flat parent record without relational child tables or JOIN operations:

# Example Schema Definition (Order Table):
repeat_ids   => 12,    # Aggregated item IDs block (e.g. "101,102,103")
repeat_start => 15,    # Repeating child rows start at block 15
blocks => [
    { id => "id",         name => "Order ID",      type => "auto_id", input => "hidden" },  # 0
    # ... fixed header fields (date, customer, address) ...
    { id => "prod_ids",   name => "Product IDs",   type => "text",    input => "hidden" },  # 12 (repeat_ids target)
    # ...
    { id => "products",   name => "Order Items",   type => "repeat",  input => "repeats" }, # 15 (repeat_start template)
];

# In-Memory Record Layout:
# $record[0]  = 1001;               # Order Primary ID (Numeric primary key)
# $record[12] = "101,102,103";      # Auto-populated by engine via repeat_fields
# $record[15] = [ 101, 'Book', 2, '150.00' ];  # 1st Product Row
# $record[16] = [ 102, 'Pad', 1, '85.00' ];    # 2nd Product Row
# $record[17] = [ 103, 'Pen', 5, '20.00' ];    # 3rd Product Row

9.7.5 Automated Validation Rules (valid)

Multiple validation rules can be chained using semicolon (;) (e.g. valid => "not_null;email"):

Rule (valid) Description Validation Check
none No Validation Field accepts any input without validation (default).
not_null Required Field cannot be null, undefined, or empty string.
unique Unique Value Asserts that no other record in the table contains this value.
email Email Format Validates RFC-compliant email pattern.
telefon Phone Number Validates national/international phone format.
ascii ASCII Only Restricts character set strictly to ASCII [0-127].
numeric Numeric Only Enforces that value is a valid numeric scalar.
regex Regular Expression Tests against custom regex pattern rule.
auto_num Auto Number Automatically assigns an incrementing numerical sequence.
auto_pass Auto Password Generates random secure password and stores salted hash.
auto_date Auto Date Automatically populates with current system timestamp.
auto_str Template String Pre-populates predefined template text.

9.7.6 Unique Constraints & Bidirectional String/ID Dictionary (.unq)

AmberDB uses .unq (Unique & Dictionary) index files (${table}_${block}.unq) to manage both uniqueness validation and relational string $\leftrightarrow$ numeric ID translation with $O(1)$ disk lookup speed:

  1. Extension Clarity: Renamed from legacy .str to .unq to eliminate any visual ambiguity with .srt (Sort indexes).
  2. $O(1)$ Duplicate Enforcement (valid => "unique"):
    • When valid => "unique" is specified (e.g. username, email, barcode), insert_id and modify_id perform an instantaneous $O(1)$ check on s:$value in ${table}_${blk}.unq.
    • If another record holds this value, the transaction is rejected with a unique constraint error.
    • Successfully written records store bidirectional mappings (s:$value => $rid and n:$rid => $value), which are automatically cleaned up when records are deleted.
  3. RDBM & match_block String-to-ID Auto-Resolution:
    • When a string name is passed to a relational field (e.g. "Can Publishing" for rdbm => "catalog_brand;1"), AmberDB queries s:Can Publishing in catalog_brand_1.unq.
    • If present, it resolves to the existing numeric ID; if absent in write mode, it auto-registers the entry in .unq and the foreign table with an incremented ID.
    • The inverted filter index (.fld) always stores pure numeric IDs, ensuring lightweight index storage and fast integer comparisons.

9.8 Schema-Driven Type Validation & Casting (enc_validate & dec_validate)

AmberDB enforces two-way data integrity between Perl runtime types and database storage:

  1. Write Phase Validation (enc_validate):
    • Invoked in insert_id, modify_id, insert_list, and modify_list right before records are written to disk and secondary indexes.
    • Cleans numeric fields, trims whitespace, and converts empty/invalid inputs to 0.
    • Normalizes non-ASCII characters for ascii fields using to_ascii.
    • Fills empty valid => "auto_date" fields with the current ISO date.
    • Converts comma-separated strings into Perl ARRAY refs for array fields and enforces HASH refs for hash fields.
  2. Read Phase Casting (dec_validate):
    • Invoked in read_id, read_list, and read_all immediately after db_decode.
    • Casts numeric fields to numeric scalars (0 + $val), eliminating uninitialized value warnings in mathematical expressions.
    • Guarantees array fields return [] and hash fields return {} even when empty.
  3. Simple Mode Performance:
    • In Schemaless (simple => 1) mode or for tables without block definitions, enc_validate and dec_validate return input data immediately with zero CPU overhead.

9.8.1 How Schemas Coordinate with CRUD Operations

When a record is added or modified via insert_id or modify_id, the passed array elements map directly to block indices:

# Block Mapping:
# Block 0 : ID (PrimaryKey - auto-generated by the engine or passed as 0)
# Block 1 : @record[0] -> Category ID ("5")
# Block 2 : @record[1] -> Brand ID ("12")
# Block 3 : @record[2] -> Author ID ("")
# Block 4 : @record[3] -> Title ("Wireless Headphones")
# Block 5 : @record[4] -> Subtitle ("Active Noise Cancelling")
# Block 6 : @record[5] -> Supplier ("Sony")
# Block 7 : @record[6] -> Description ("<p>Detailed product description...</p>")
# Block 8 : @record[7] -> Stock ("150")
# Block 9 : @record[8] -> Barcode ("8690123456789")
# Block 10: @record[9] -> Price ("2499.90")
# Block 11: @record[10]-> Status ("1")

my @product = (
    "5", "12", "", "Wireless Headphones", "Active Noise Cancelling",
    "Sony", "<p>Detailed product description...</p>", 150, "8690123456789", 2499.90, "1"
);

my $new_id = $adb->insert_id("catalog_product", 0, @product);

In a single atomic pass, the engine consults the schema and:

  1. Validates and normalizes field types via enc_validate.
  2. Writes the raw record to catalog_product.db.
  3. Updates catalog_product.inx primary index (since record_index => 1).
  4. Indexes Category (5), Brand (12), and Status (1) in catalog_product_*.fld match indexes (since match_block => [1, 2, 3, 11]).
  5. Extracts, tokenizes, normalizes, and indexes Title, Subtitle, Description, and Barcode in catalog_product_*.src inverted search indexes.
  6. Generates the URL slug sony-wireless-headphones into catalog_product.slg (since slug_block => [2, 4]).

9.9 Dynamic Runtime Schema Manipulation (table_attr)

AmberDB schemas are mutable at runtime without database recreation or migrations:

# Scenario 1: Narrow full-text search scope dynamically for barcode POS scanners
$adb->table_attr("catalog_product", { search_block => [ 4, 9 ] });

# Scenario 2: Include soft-deleted records or enable audit logging dynamically
$adb->table_attr("catalog_product", { keep_deleted => 1 });

# Scenario 3: Temporarily disable cache during heavy batch ETL or reporting
$adb->table_attr("catalog_product", { use_cache => 0 });

9.10 Dynamic Expanding Tables and Repeating Blocks (repeat_ids & repeat_start)

AmberDB breaks free from fixed column width constraints by allowing a variable number of child items (e.g. order line items, cart items, invoice rows) to be appended dynamically at the end of a single parent document record. This feature eliminates child junction tables (orders $\leftrightarrow$ order_items) and multi-table SQL JOIN operations entirely.

9.10.1 Schema Configuration (order_active.table Example)

# dbstore/schema/order_active.table
{
    name         => "Active Orders",
    record_index => 1,
    match_block  => [ 1, 2, 12, 14 ],   # 12: Product Loop (repeat_ids) is automatically indexed
    keep_deleted => 1,
    log_owner    => 1,
    repeat_ids   => 12,                 # Block where extracted child IDs are consolidated
    repeat_start => 15,                 # Starting index where variable child blocks begin

    blocks => [
        { id => "id",                name => "ID",                   type => "auto_id" }, # 0
        { id => "member_id",         name => "Member ID",            type => "text" },    # 1
        { id => "invoice_no",        name => "Invoice No",           type => "text" },    # 2
        { id => "amounts",           name => "Amounts",              type => "array" },   # 3
        { id => "timestamps",        name => "Timestamps",           type => "array" },   # 4
        { id => "status",            name => "Status",               type => "option" },  # 5
        { id => "session_id",        name => "Session ID",           type => "text" },    # 6
        { id => "delivery_address",  name => "Delivery Address",     type => "array" },   # 7
        { id => "invoice_address",   name => "Invoice Address",      type => "array" },   # 8
        { id => "cargo",             name => "Shipping Info",        type => "array" },   # 9
        { id => "payment_info",      name => "Payment Method",       type => "array" },   # 10
        { id => "credit_card_info",  name => "Card Info",            type => "array" },   # 11
        { id => "product_ids",       name => "Product Loop",         type => "text" },    # 12 (repeat_ids)
        { id => "member_notes",      name => "Customer Notes",       type => "array" },   # 13
        { id => "gift_products",     name => "Gift Products",        type => "text" },    # 14
        { id => "products",          name => "Order Items",          type => "repeat" },  # 15 (repeat_start)
    ]
}

9.10.2 Engine Processing & Automatic Indexing (repeat_fields)

During every insert_id, modify_id, insert_list, or modify_list call, the engine automatically processes all repeating blocks starting from repeat_start (15):

  1. It extracts the identifier of each repeating block (the first element $_->[0] if it’s an ARRAY reference, or the scalar value itself).
  2. It joins these IDs into a comma-separated string ("101,102,103") and assigns it automatically to block repeat_ids (12) — developers do not need to populate this field manually.
  3. Because Block 12 is declared in match_block, the engine automatically indexes each product key into order_active_12.fld via field_to_list.

[!NOTE] Repeating Blocks in Schemaless Simple Mode:
Automatic compilation and comma-joining of repeating child IDs into repeat_ids (repeat_fields) relies strictly on the repeat_start and repeat_ids attributes in the .table schema file. In schemaless Simple Mode (simple => 1), schema directives are inactive and this automatic aggregation does not execute; records are written as raw Perl arrays. If a summary ID list is needed in Simple Mode, it must be populated manually by the developer before writing.

9.10.3 Code Example & Direct Querying

# 1. Insert Order with Expanding Product Items (Starting at Block 15)
my @order = (
    "1001",              # [1] Member ID
    "INV-2026-001",      # [2] Invoice No
    "2199.00",           # [3] Total Amount
    "2026-08-24",        # [4] Order Date
    "1",                 # [5] Status (Active / Confirmed)
    "SESS12345",         # [6] Session
    "Delivery Address",  # [7] Delivery
    "Invoice Address",   # [8] Invoice
    "Shipping ID",       # [9] Shipping ID
    "CreditCard",        # [10] Payment
    "**** 1234",         # [11] Card
    "",                  # [12] product_ids (Leave empty; engine fills with "101,102,103")
    "Ring bell",         # [13] Notes
    "Gift Wrap",         # [14] Gift
    [ "101", "MacBook Pro M3", 1, 1999.00 ], # [15] Product 1 (repeat_start)
    [ "102", "Magic Mouse",    2,   99.00 ], # [16] Product 2
    [ "103", "USB-C Hub",      1,   49.00 ], # [17] Product 3
);

my $order_id = $adb->insert_id("order_active", undef, @order);

# 2. Query ALL Active Orders containing Product 101 via direct key seek:
my @orders = $adb->field_fetch("order_active", 12, "101");
print "Found " . scalar(@orders) . " active orders containing product 101.\n";

9.11 Vertical Partitioning & Child Tables (parent_table)

For scenarios involving very large or infrequently accessed data blocks (such as rich HTML descriptions, technical sheets, or multi-paragraph document bodies), keeping the primary table’s record footprint compact maximizes search and index caching speeds. AmberDB natively supports Vertical Partitioning:

# dbstore/schema/catalog_descript.table
{
    name         => "Product Descriptions",
    parent_table => "catalog_product",
    blocks => [
        { id => "id",          name => "ID",          type => "auto_id" }, # 0 (Shares Product ID)
        { id => "description", name => "HTML Content",type => "text" },    # 1
    ]
}

Architectural Advantage:

  1. Category listings and search queries stream lightweight product records without loading megabytes of rich HTML descriptions into memory.
  2. Only when a visitor navigates to a specific product detail page is $adb->read_id("catalog_descript", $product_id) called to fetch the full rich content in a single direct key seek.

10. Database Group Structure (.dbase)

To group related tables and apply automated partitioning (by year or branch), define a .dbase file:

# dbstore/schema/catalog.dbase
{
    name    => "Catalog Database Group",
    type    => 0,                           # 0: System table, 1: Dynamic table
    year    => 0,                           # 1: Partition into yearly folders (e.g. 2026/invoice.db)
    section => 0,                           # 1: Partition by branch/section
};

11. Smart Tiered (Hot / Cold Junk) Indexing

Over time, hundreds of thousands of products go out of stock, become discontinued, or vendor contracts end. You cannot delete these records (they must remain intact for order history, invoices, and accounting), but they should never slow down active customer search or category browsing.

The Junk Subsystem is an automated performance shield that partitions your data into Active (Storefront) and Junk (Archive) tiers without any data loss.

11.1 Key Benefits & Features

11.2 Schema Configuration (.table)

Enable dual-tier indexing and declare your business rules in junk_rules:

# dbstore/schema/catalog_product.table
{
    name         => "Products",
    record_index => 1,
    use_junk     => 1,                       # Enables smart hot/cold indexing
    
    # Define conditions that qualify a record as "Junk / Archive":
    junk_rules   => [
        # 1. Product's own sales status (Block 20) is not 1 (Active) -> ARCHIVE
        [ 20, "ne", 1 ],

        # 2. Relational Vendor Rule: Publisher (Block 2) status is disabled in catalog_producer -> ARCHIVE
        [ "2->14", "ne", 1 ],
    ],
    
    jnktype      => "AB",                    # Default query mode (Active first, then archive)
    search_block => [ 4, 5 ],
    match_block  => [ 1, 2, 3 ],
}

11.3 Usage Scenarios & Code Examples

Select the optimal query tier using the jnktype parameter:

A. Storefront & Category Pages (Active Items Only - Mode A)

Keep category listings and customer browsing clean of obsolete items:

# Read active products for category listing:
my @storefront_items = $adb->read_all("catalog_product", jnktype => "A");

# Customer search:
my @results = $adb->search_table("catalog_product", "headphones", jnktype => "A");

B. Storewide Search (Active First, Archived Items Appended - Mode AB)

Ensure rare or older items remain discoverable without burying in-stock products:

# Active products rank first, discontinued items appear at the end:
my ($total, @results) = $adb->search_table("catalog_product", "clean code", 0, 20, jnktype => "AB");

C. Back-Office Admin & Reports (Archived Items Only - Mode B)

Inspect discontinued, out-of-stock, or passive catalog items:

# List all archived/junk product IDs:
my @archived_ids = $adb->read_all("catalog_product", jnktype => "B", keys_only => 1);

D. Order & Invoice Processing (Direct ID Access)

Past orders access product details seamlessly regardless of whether the item is active or archived:

# Fetch product details directly by ID (Works instantly for both active and archived products):
my @product = $adb->read_id("catalog_product", $old_product_id);

11.4 Automatic State Migration

When updating a product, AmberDB evaluates the schema rules in real time:

12. Automated URL Slug Management

When slug_block => [2, 4] is configured (Brand + Title), AmberDB generates and manages clean URL slugs automatically:

# Retrieve URL Slug by Record ID
my $slug_map = $adb->get_slug("catalog_product", 0, 5001);
my $slug    = $slug_map->{5001};
print "URL: /product/$slug\n"; # Output: /product/acme-wireless-headphones

# Resolve Record ID from URL Slug (Router lookup)
my $id_map = $adb->get_slug("catalog_product", 1, "acme-wireless-headphones");
my $id     = $id_map->{"acme-wireless-headphones"};
print "Resolved Product ID: $id\n";

12.1 Automatic Slug Collision Resolution (Numeric Suffixes)

When multiple records generate identical base slugs (e.g. two distinct products named “Wireless Headphones”), AmberDB automatically appends deterministic incrementing numeric suffixes (_2, _3) to ensure strict uniqueness:

13. Unified Shared RAM Cache (.db / .inx) & Persistent Buffer

AmberDB::Cache provides a unified shared RAM cache mirroring AmberDB’s native .db and .inx formats:

                               ┌────────────────────────────────────────────────┐
                               │       dbstore/cache/ (tmpfs RAM-Disk)          │
                               ├──────────────────────┬─────────────────────────┤
                               │ cache/${table}.db    │ cache/${table}.inx      │
                               │ (Records)            │ (lastid, keys, meta...) │
                               └──────────────────────┴─────────────────────────┘

Cache Levels (use_cache)

# 1. Manual Cache Write (stores in cache/${table}.inx)
$adb->cache_write("catalog_product", "featured_items", @featured_list);

# 2. Cache Read
my @featured = $adb->cache_read("catalog_product", "featured_items");

# 3. Hard Cache Table Preload
$adb->cache_preload("catalog_category");

# 4. Invalidate Cache (Automatically purged on modify / delete_id)
$adb->cache_delete("catalog_product", "featured_items"); # Single key
$adb->cache_delete("catalog_product");                   # Entire table cache (.db and .inx)

# 5. Inspect RAM-Disk Diagnostics & Mount Status
my $cache_diag = $adb->cache_setup();
# Returns hashref: { is_mounted => 1, mount_desc => "...", cache_dir => "...", cache_size => "512M" }

Cache TTL (cache_ttl) & Runtime Overrides

The cache_ttl expiration time is defined per-table directly inside its schema (e.g. cache_ttl => 1800). Ephemeral data structures like session tokens or process locks can have their expiration configured in the schema or dynamically tuned at runtime using table_attr:

# Dynamically configure session table cache TTL to 30 minutes (1800 seconds)
$adb->table_attr("session", { use_cache => 1, cache_ttl => 1800 });

Temporary Disk Buffer

For large reporting queries or intermediate batch jobs:

$adb->buffer_write("temp_report", @large_data);
my @data = $adb->buffer_read("temp_report");
$adb->buffer_delete("temp_report");

14. Configuration and Deterministic Flag Management (config)

Runtime behavior can be tuned and safely configured via the $adb->config() method:

# Bulk or single configuration assignment (Recommended)
$adb->config(
    no_write   => 1,              # Read-only maintenance mode: block all writes
    no_backup  => 1,              # Disable daily CSV audit logging for all tables
    simple     => 1,              # Direct unindexed mode: bypasses secondary index generation
    keys_only  => 1,              # read_all returns IDs only
    cache_size => '1024M',        # RAM-Disk / tmpfs cache size (Default: 512M)
);

# Single scalar getter:
my $no_write = $adb->config('no_write');

# Bulk getter (returns a safe shallow copy):
my $cfg = $adb->config();

15. Data Structures, Low-Level Table and Stream Operations

Beneath the standard CRUD layer, AmberDB provides direct access to optimized DB_File C-level primitives and raw streaming methods:

15.1 Data Structures and Serialization (db_encode, db_decode)

AmberDB encodes and decodes complex nested Perl structures:

# Encode: Native Perl Data → String
my $encoded = $adb->db_encode("Text", [ 1, 2, 3 ], { key => "val" });

# Decode: String → Native Perl Data
my ($text, $arr_ref, $hash_ref) = $adb->db_decode($encoded);

15.2 Low-Level Table and Stream Management (table_read, table_write, table_close)

Used for direct batch processing sessions or custom streaming tasks:

my $table_path = $adb->table_path("catalog_product") . ".db";

# 1. Open Table in Read/Write Mode with Exclusive Lock (flock LOCK_EX)
my $db_obj = $adb->table_write($table_path);

# 2. Open Table in Read-Only Mode (O_RDONLY)
my $db_ro  = $adb->table_read($table_path);

# 3. Synchronize (sync), Unlock, and Close Table Session
$adb->table_close($table_path);

15.3 Raw Record Manipulation (recs_get, recs_put, recs_del, recs_exist, recs_keys, recs_scan, table_readid)

Executes direct $db->get(), $db->put(), and $db->del() calls on open or dynamically resolved table handles:

# 1. Bulk Read Raw Values (recs_get)
my $raw_data = $adb->recs_get($table_path, 5001, 5002);
# Returns: { 5001 => "raw_encoded_string", 5002 => "..." }

# 2. Single Record Direct Read with Auto-Session (table_readid)
my ($rid, @record) = $adb->table_readid($table_path, 5001);

# 3. Bulk Put Raw Records (recs_put)
$adb->recs_put($table_path, 
    [ 5001, "5,12", "3", "7", "Product A", "", "", "", "", "199.00", "1" ],
    [ 5002, "5",    "8", "9", "Product B", "", "", "", "", "299.00", "1" ]
);

# 4. Check Key Existence (recs_exist)
my $exists = $adb->recs_exist($table_path, 5001);

# 5. Retrieve All Raw Keys from Open Table (recs_keys)
my @keys = $adb->recs_keys($table_path);

# 6. Stream/Iterate Over All Records without High Memory Overhead (recs_scan)
$adb->recs_scan($table_path, sub {
    my ($key, $raw_val) = @_;
    # Process record stream lazily
});

# 7. Bulk Delete Raw Records (recs_del)
$adb->recs_del($table_path, 5001, 5002);

15.4 Table Metadata and ID Helpers (table_keys, table_count, table_lastid, table_autoid, table_create)

# Retrieve array of all active primary keys
my @all_ids = $adb->table_keys("catalog_product");

# Total active record count
my $total = $adb->table_count("catalog_product");

# Highest (latest) primary key
my $last_id = $adb->table_lastid("catalog_product");

# Generate or format next auto-increment ID
my $new_id = $adb->table_autoid("catalog_product");

# Initialize empty .db file for table
$adb->table_create("catalog_product");

15.5 String & Text Processing Utilities (AmberDB::String)

Since AmberDB inherits from AmberDB::String, a suite of fast string sanitization, formatting, and classification helpers are directly accessible on $adb:

# 1. Whitespace Normalization & Flattener (trim_space)
my $clean = $adb->trim_space("  hello \n\t world  ");      # Preserves line breaks
my $flat  = $adb->trim_space("  hello \n\t world  ", 1);   # Flattens all whitespace to single space

# 2. HTML Tag Stripping (remove_tags)
my $text = $adb->remove_tags("<p>Description with <br/>line break</p>");

# 3. Text Truncation with Ellipsis Preservation (truncate_text / sub_str / short_title)
my $summary = $adb->truncate_text($long_body, 120);        # Word-boundary safe truncation
my $short   = $adb->short_title($product_title, 32);       # ASCII-normalized short slug/title

# 4. Data Pattern Classifier (what_isthis)
my $type = $adb->what_isthis("user@example.com");          # Returns: 'email'
# Recognizes: email, barcode, gsm, phone, tcno, number, ascii, letter, domain, other

# 5. HTML Entity Conversion (html_ascode / code_ashtml / text2html / html2text)
my $encoded_html = $adb->html_ascode('<a href="test">');   # Encodes special characters to HTML entities
my $plain_text   = $adb->html2text($html_document);

16. Faceted Search & Category Filters (Facet Engine)

The Facet Engine powers e-commerce sidebar filter menus (Brand, Category, Author, Price Range, Color, etc.), designed for high-performance, low-latency multi-select faceted filtering across large product catalogs.

16.1 Key Benefits & Features

16.2 Schema Configuration (.table)

Enable the facet engine by adding use_facet => 1 and your facet_block specifications to your table schema:

# dbstore/schema/catalog_attributes.table
{
    name         => "Product Attributes",
    use_facet    => 1,                       # Enables the facet filtering engine on this table
    
    # Define which blocks to expose as sidebar filters:
    facet_block  => [
        # Relational Filters (Category, Brand, Author from foreign tables):
        { blk => 1, id => "category", label => "Category",    table => "catalog_category",    name_idx => 2 },
        { blk => 2, id => "brand",    label => "Brand",       table => "catalog_producer",    name_idx => 2 },
        { blk => 3, id => "author",   label => "Author",      table => "catalog_contributor", name_idx => 2 },
        
        # Numeric / Range Filters:
        { blk => 4, id => "price",    label => "Price Range" },
        
        # Free-Text Attributes (Color, Size, etc.):
        { blk => 6, id => "color",    label => "Color" },
    ],
}

16.3 Usage & Practical Examples

A. Building Category Sidebar Menus

Generate complete filter groups and matching product counts in a single method call:

# User selections from URL query string: Category 5, Brand 12 or 14 selected
my %selected_filters = ( 1 => "5", 2 => ["12", "14"] );

my $menu = $adb->facet_menu(
    "catalog_attributes",
    \%selected_filters,
    $table_info->{facet_block},
    { limit => 10, sort => "count" } # Display top 10 options per group sorted by product count
);

# $menu structure is ready to pass directly to your template:
# {
#     count         => 42,                         # Total matching products
#     ids           => [ 101, 105, 120, ... ],     # IDs of matching products for product grid
#     active_counts => { 1 => 1, 2 => 2 },         # Active filters count per block
#     groups        => [                           # Ready-to-render sidebar groups:
#         {
#             blk          => 2,
#             name         => "Brand",
#             active       => "1",
#             active_count => 2,
#             records      => [
#                 { uid => "fc_2_12", param => "f2", val => 12, label => "Apple",   count => 28, checked => "1" },
#                 { uid => "fc_2_14", param => "f2", val => 14, label => "Samsung", count => 14, checked => "1" },
#                 { uid => "fc_2_19", param => "f2", val => 19, label => "Sony",    count => 6,  checked => ""  },
#             ]
#         },
#         ...
#     ]
# }

B. Dynamic Filters on Search Result Pages

Pass the list of search result IDs as base_ids so sidebar filters apply strictly to search results:

# 1. Search catalog for user query (keys_only returns unpaginated ID list)
my @found_ids = $adb->search_table("catalog_product", "sci-fi", keys_only => 1);

# 2. Generate facet menu scoped exclusively to the search results
my $search_facets = $adb->facet_menu(
    "catalog_attributes",
    \%selected_filters,
    $table_info->{facet_block},
    { base_ids => \@found_ids }
);

17. User Audit Trail and Backup

17.1 User Action History (log_owner)

When log_owner => 1 is enabled in the schema, record modification history is stored in .aut:

# Retrieve user audit history as formatted HTML
my $history_html = $adb->auth_view("catalog_product", 5001);
print $history_html;
# Output:
#     add     2026-08-14 10:15    admin_user
#     edit    2026-08-14 11:30    editor_user

17.2 Continuous Recovery Stream (YYYY-MM-DD.csv)

AmberDB automatically appends every insert, modify, and delete operation into a clean, chronological time-series stream in backup/YYYY/YYYY-MM-DD.csv.

Each entry is tab-separated (\t) using the standard format: [Timestamp] \t [User] \t [Action] \t [Table] \t [Record ID] \t [Packed Values]

To disable this backup stream:

17.3 Native Database Archive (.amberdb Dump & Restore)

AmberDB packages all schemas (schema/*.table, schema/*.dbase) and authoritative data files (tables/*.db, tables/*.del, tables/*.aut, tables/*.cnt) alongside cryptographically verified SHA-256 checksums in a single compressed, portable .amberdb archive file that mirrors the native physical database directory structure.

Derived index files (.inx, .src, .fld, .fac, .srt) are intentionally excluded to keep archives compact and ensure future-proof portability; restore deterministically rebuilds all indexes via set_index.

use AmberDB;
use AmberDB::Tools;

my $adb   = AmberDB->new(path => { dbase_dir => "./dbstore" });
my $tools = AmberDB::Tools->new($adb);

# 1. Create full database backup archive (.amberdb)
my $archive = $tools->dump();
# Output: dbstore/backup/2026/amberdb_2026-08-28_180000.amberdb

# 2. Export specific tables as a focused snapshot archive
$tools->dump(
    file   => "backup/2026/catalog_backup.amberdb",
    tables => ["catalog_product", "catalog_category"]
);

# 3. Restore database archive and automatically rebuild all indexes
$tools->restore(
    file    => "backup/2026/catalog_backup.amberdb",
    force   => 1, # Overwrite confirmation for non-empty target directories
    reindex => 1  # Automatically reconstruct binary indexes from source data
);

CLI Command-Line Utility (bin/amberdb_backup.pl)

# Dump entire database to default archive
perl bin/amberdb_backup.pl --dump --file backup/2026/full_backup.amberdb

# Dump specific tables only
perl bin/amberdb_backup.pl --dump --tables products,orders

# Restore database archive with integrity checks and automatic reindexing
perl bin/amberdb_backup.pl --restore --file backup/2026/full_backup.amberdb --force

18. Maintenance and Repair Tools (AmberDB::Tools)

AmberDB::Tools provides utilities for reindexing, table vacuuming, and data migration:

use AmberDB;
use AmberDB::Tools;

my $adb   = AmberDB->new(path => { dbase_dir => "./dbstore" });
my $tools = AmberDB::Tools->new($adb);

# 1. Rebuild all indexes for a table
$tools->set_index("catalog_product");

# 2. Rebuild indexes across all tables in database
$tools->index_alltables();

# 3. Verify index consistency
my @records = $adb->read_all("catalog_product", 0, 0, no_index => 1);
my $diff    = $tools->check_readall("catalog_product", @records);

# 4. Vacuum Table (Removes fragmentation and shrinks .db file)
$tools->vacuum("catalog_product", 1); # 1 = automatically reindex after vacuum

# 5. Export / Import CSV
$tools->tie2csv("catalog_product");
$tools->csv2tie("catalog_product");

# 6. Batch Reindex / Convert All Database Tables
my $converted_report = $tools->convert_tables();

# 7. Delete Table and All Secondary Index Files from Disk
$tools->del_table("obsolete_table");

# 8. Lightweight Ad-Hoc AmberDB Instance for Temporary/Standalone Dirs
my $simple_adb = $tools->db_simple("/path/to/data/dir");

19. File Extensions Map

AmberDB file extensions are classified into 3 operational tiers based on their authority and reconstructibility:

Extension Role / Classification Reconstructible? Description
Authoritative Master Data      
.db Primary Data (Source of Truth) ❌ No (Authoritative) Berkeley DB master document table (DB_File Hash).
.del Soft-Deleted Archive ❌ No (Authoritative) Archive of soft-deleted records (keep_deleted).
.aut User Audit Trail ❌ No (Authoritative) Chronological user action log (log_owner).
.str String Dictionary Mapping ❌ No (Authoritative) Bidirectional string-to-foreign-key dictionary file (_${blk}.str).
Derived Secondary Indexes      
.inx Record Index Yes (set_index) Binary array of all active IDs, total count, highest ID.
.fld Inverted Match Index Yes (set_index) Block-level key-to-IDs inverted index (match_block).
.src Full-Text Search Index Yes (set_index) Word-level token inverted index (search_block).
.srt Sort Index Yes (set_index) Pre-sorted binary array of record IDs (sort_block).
.fac Facet Navigation Index Yes (set_index) Forward index for faceted filter navigation (facet_block).
.slg URL Slug Map Yes (set_index) Bidirectional map: _0.slg (ID→Slug) and _1.slg (Slug→ID).
.jinx Junk Record Index Yes (set_index) Binary primary index for cold/archived records (use_junk).
.jfld Junk Match Index Yes (set_index) Field match index for cold records (jnktype => 'B'/'AB').
.jsrc Junk Full-Text Search Yes (set_index) Word-level inverted index for cold records (jnktype => 'B'/'AB').
Runtime & Transient Files      
.cnt View / Hit Counter ⚠️ Counter state Hit/read counter file (use_counter).
.txn Transaction Undo Journal ⚠️ Transient (Runtime) Active transaction rollback journal file (txn/).
.cache Shared RAM-Disk Cache Yes (RAM-Disk) RAM-Disk shared cache file (cache/).
.tmp Disk Buffer File ⚠️ Transient (Staging) Disk staging buffer file under dbstore/buffer/ (buffer_write).
.lock Process Mutex Lock ⚠️ Transient (Mutex) OS flock process synchronization lock file.

20. Directory Structure

dbstore/
├── schema/                      ← Schema and Group Configurations
│   ├── catalog.dbase            ← Group definition
│   ├── catalog_product.table    ← Product table schema
│   └── catalog_category.table   ← Category table schema
├── tables/                      ← Main Data and Index Files
│   ├── catalog_product.db       ← Main data file
│   ├── catalog_product.inx      ← Binary record index
│   ├── catalog_product_1.fld    ← Category match index
│   ├── catalog_product_4.src    ← Title search index
│   ├── catalog_product_10.srt   ← Price sort index
│   ├── catalog_product.fac      ← Facet index
│   ├── catalog_product_0.slg    ← ID → Slug Map
│   ├── catalog_product_1.slg    ← Slug → ID Map
│   ├── catalog_product.aut      ← Audit trail
│   └── catalog_product.del      ← Soft-deleted records
├── cache/                       ← Shared RAM-Disk Cache Files
├── buffer/                      ← Transient Disk Buffer / Staging Files
├── txn/                         ← Active Transaction Journals
├── pids/                        ← Lock Files
└── backup/                      ← Daily CSV Backups

21. Developer Best Practices and Recommendations

  1. Use insert_list for Bulk Ingestion: When adding hundreds of records, use insert_list instead of looping over insert_id. Batch mode writes all records in a single file session and rebuilds indexes in one pass.
  2. Wrap Multi-Step Writes in transact_start: Always wrap inventory deductions, checkout sequences, or multi-table balance updates inside transactions.
  3. Index Only Required Fields: Only assign fields to match_block or search_block if they are actively queried to minimize disk write overhead.
  4. Always Handle Pagination Return Signatures Correctly: When passing $limit > 0 to read_all, field_fetch, or search_table, remember that the first returned value is $total_count integer. Never unpack into a single array (my @records = $adb->read_all(..., 0, 20)) as $records[0] will be an integer scalar causing fatal crashes upon dereferencing. Always unpack paginated queries as my ($total_count, @records).
  5. Choose Numeric IDs Where Possible: Standardize on id_type => "num" for optimal 64-bit binary packing performance.
  6. Standardize on Record Array ID at Index 0: Always maintain the Primary Key ID at Index 0 ($record[0]) within record arrays (@record). For new records, initialize with 0 and assign the returned ID via my $id = $record[0] = $adb->insert_id("table", @record);. Performing retrieval (read_id), updating (modify_id("table", @record)), and deletion (delete_id("table", $record[0])) against this unified structure ensures clean code and eliminates positional argument shifting bugs.

22. Full Working Example (Checkout & Stock Transaction Scenario)

The following example demonstrates creating master entity tables, inserting a product with referenced foreign IDs and multi-category indexing, querying with sorting, and executing an atomic checkout transaction:

use strict;
use warnings;
use AmberDB;

# 1. Initialize Engine
my $adb = AmberDB->new(
    cfg  => { language => "en", user => "cashier_1" },
    path => { dbase_dir => "./dbstore" }
);

# 2. Populate Master Entity Tables
my $cat_computers = $adb->insert_id("catalog_category", undef, "Computers & IT", 1); # ID: 5
my $cat_portable  = $adb->insert_id("catalog_category", undef, "Portable Devices", 1);# ID: 12

my $brand_apple   = $adb->insert_id("catalog_brand", undef, "Apple", "USA");          # ID: 8
my $author_team   = $adb->insert_id("catalog_author", undef, "Hardware R&D", "Core"); # ID: 7

# 3. Add New Product (Relational fields receive IDs; multi-category stored as "5,12")
my @product = (
    "5,12",                         # [1] Category IDs (5: Computers, 12: Portable)
    "8",                            # [2] Brand ID: Apple (8)
    "7",                            # [3] Author / Contributor ID: 7
    "MacBook Pro M3",               # [4] Product Title
    "16GB RAM 512GB SSD Space Gray",# [5] Subtitle
    "", "", "",
    10,                             # [8] Stock Count: 10 units
    "195949123456",                 # [9] Barcode
    "1999.00",                      # [10] Price
    "1"                             # [11] Status: Active
);

my $product_id = $adb->insert_id("catalog_product", undef, @product);
print "1. Product created -> ID: $product_id\n";

# 4. Read Auto-Generated URL Slug
my $slug_map = $adb->get_slug("catalog_product", 0, $product_id);
print "2. Product URL -> /product/$slug_map->{$product_id}\n";

# 5. Query Multi-Category (e.g. Category 12) Sorted by Price
my ($total, @items) = $adb->field_fetch(
    "catalog_product", 1, "12", 0, 10,
    sort => { blk => 10, reverse => 1 } # Price ascending
);
print "3. Listed $total products in Category 12.\n";

# 6. Atomic Checkout Transaction
$adb->transact_start();

my @current = $adb->read_id("catalog_product", $product_id);
if ($current[8] >= 1) { # Check available inventory
    # Deduct 1 unit (@current[0] contains $product_id)
    $current[8] -= 1;
    $adb->modify_id("catalog_product", @current);
    
    # Create order (Items stored as nested ARRAY in Block 3)
    my @order_items = ( [ $product_id, "MacBook Pro M3", 1, 1999.00 ] );
    my $order_id = $adb->insert_id("orders", undef, "Customer John", time(), \@order_items, { status => "confirmed" });
    
    my $txn = $adb->transact_end();
    if ($txn->{status} eq "commit") {
        print "4. Order #$order_id placed! Remaining stock: $current[8]\n";
    }
} else {
    $adb->transact_rollback();
    print "4. Error: Out of stock! Transaction rolled back.\n";
}

23. Method Quick Reference Table

Method Arguments Return Value Description
Core CRUD Operations      
insert_id $table, $id, @fields $new_id Inserts single record and updates all indexes.
insert_list $table, @records \%status High-throughput bulk insert (bypasses txn log).
modify_id $table, $id, @fields 1/undef Updates record and synchronizes indexes.
modify_list $table, @records \%status High-throughput bulk update.
delete_id $table, $id 1/undef Deletes record (or moves to .del soft-delete).
delete_list $table, @ids \%status High-throughput bulk delete.
Reading and Querying      
read_id $table, $id @fields Reads single record by primary key ID.
read_all $table, [$s, $l, %opt] ($count, @records) Paginated & sorted read of all records.
read_list $table, \@id_list @records Reads records in given ID order.
field_fetch $table, $blk, $val, [%opt] ($count, @records) (paginated) / @records Direct key lookup via inverted match index (.fld).
search_table $table, $query, [%opt] ($count, @records) (paginated) / @records Full-text keyword search via search index.
field_filter $table, \%filter_opts { count, ids } Multi-block composite query with sorting.
field_fltkeys $table, \%facet_opts \%counts Computes dynamic facet count maps.
Existence & Positional Lookups      
exist_id $table, $id 1/0 Checks whether a record ID exists in table.
exist_list $table, @ids \%status Returns presence map { id => 1/0 } for multiple IDs.
exist_table $table, [$ext] 1/0 Checks whether physical table/index file exists.
read_firstid $table @fields Reads first record by ascending numeric ID.
read_lastid $table @fields Reads latest record by descending numeric ID.
read_randid $table @fields Reads a random record from table.
read_count $table, $id $count Reads hit/view counter from .cnt file.
Low-Level Table & Stream I/O      
table_read $file_path $db_obj Opens DB_File handle in read-only mode (O_RDONLY).
table_write $file_path $db_obj Opens DB_File handle in R/W mode with flock LOCK_EX.
table_close $file_path 1 Syncs DB_File, releases lock, and closes handle.
table_keys $table @ids Retrieves array of all active primary keys.
table_count $table $total Returns total active record count.
table_lastid $table $last_id Returns highest allocated primary key.
table_autoid $table, [$id] $new_id Generates or formats next auto-increment ID.
table_create $table 1 Creates empty .db table file on disk.
recs_get $file_path, @ids \%result Direct $db->get() reading { id => raw_val }.
recs_put $file_path, @records 1 Direct batch $db->put() for [$id, @fields] records.
recs_del $file_path, @ids 1 Direct batch $db->del() for provided record IDs.
recs_cutting $start, $limit, @list ($count, @slice) In-memory array pagination slicer.
Transaction Management      
transact_start — 1/undef Starts a new transaction with undo journaling.
transact_end — \%result Commits transaction or triggers auto-rollback.
transact_rollback — \%result Forces immediate manual rollback.
Cache, Slug, Schema & Audit      
table_info $table \%schema Retrieves active table schema configuration hash.
table_attr $table, \%attrs 1 Dynamically mutates in-memory table schema at runtime.
cache_read $table, $key @data Reads from RAM-Disk shared cache.
cache_write $table, $key, @data 1 Writes to RAM-Disk shared cache.
cache_delete $table, [$key] 1 Purges cache entries.
get_slug $table, $type, @keys \%map Resolves ID ↔ URL slug mappings.
auth_view $table, $id $html Returns user audit trail as HTML.

24. Why Use AmberDB? (Comparison with SQL and SQLite)

AmberDB is not designed to be a “weaker SQL engine” trying to mimic relational databases. Instead, it solves problems where relational models impose excessive complexity, joins, triggers, and boilerplate application code by leveraging native, schema-driven, unified document structures and inverted indexing.

24.1 Unified Nested Records and Eliminating SQL JOINs

In relational SQL databases (MySQL, PostgreSQL, SQLite), storing an order with multiple line items and metadata requires table normalization (orders, order_items, attributes) and complex multi-table JOIN operations during retrieval.

In AmberDB, records are stored in a unified (denormalized) native Perl structure:

my @order = (
    "Customer_A",                 # [1] Customer Name
    "2026-08-14",                 # [2] Order Date
    [                             # [3] Nested ARRAY: Order Items (Product IDs: 101, 102)
        [ 101, "Laptop", 1, 35000 ],
        [ 102, "Wireless Mouse", 2, 750 ]
    ],
    { status => "confirmed", tracking_code => "TR12345" } # [4] Nested HASH: Metadata
);

$adb->insert_id("orders", 1001, @order);

This entire document is written to the .db file as a single key-value pair. When read via $adb->read_id("orders", 1001), it is instantly returned as native Perl array and hash references ready for immediate use, completely avoiding JSON deserialization overhead or multi-table SQL joins.

24.2 Resolving Relationships with Low I/O via match_block

In SQL, answering “Which orders contain Product 101?” requires scanning the order_items index/table, joining with orders, and executing multiple disk/cache seeks across separate tables.

In AmberDB: The order record contains the array of product items in Block 3. When match_block => [3] is defined in the schema, the engine automatically extracts each product ID using field_to_list and indexes it into orders_3.fld.

# Fetch all order records containing Product 101:
my @orders = $adb->field_fetch("orders", 3, 101);

This operation executes a single direct key lookup from orders_3.fld, retrieving all Order IDs matching the key 101 directly (with O(1) average-time lookup per indexed key):

# Inside orders_3.fld:
# 101 => [ 1001, 1005, 1023 ] (Packed binary RID array)

After retrieving the keys, the engine reads their record values in a single pass and returns all detailed information belonging to the matching orders.

While SQL engines traverse multiple tables, B-Trees, and relational joins; AmberDB resolves the query directly via precomputed inverted indexes, eliminating redundant disk I/O and query-planning overhead.

24.3 Schema-Driven Automated Multi-Indexing on CRUD

In SQL, you must manually manage CREATE INDEX statements, full-text indexes, and trigger logic or application glue code to keep search indexes synchronized.

In AmberDB, you declare indexes once in the table’s .table schema file:

{
    match_block  => [1, 3],    # Customer ID & Product ID match index (.fld)
    search_block => [4],       # Full-text search index (.src)
    facet_block  => [1, 2],    # Faceted navigation index (.fac)
    sort_block   => [10],      # Binary sorted price index (.srt)
    slug_block   => [1, 4],    # Bidirectional URL slug index (.slg)
    log_owner    => 1,         # User audit trail (.aut)
    keep_deleted => 1,         # Soft-delete archive (.del)
}

Whenever you execute $adb->insert_id(...), $adb->modify_id(...), or $adb->delete_id(...), the engine automatically synchronizes the base table and all corresponding index files in one atomic step.

24.4 Direct Inverted Key Lookups (Zero Query Planner Overhead)

In SQL, running SELECT id FROM orders WHERE customer_id = 'A' requires parsing, query plan evaluation, cost optimization, and virtual machine execution.

In AmberDB, field_fetch is a direct hash key lookup on Berkeley DB returning packed binary buffers. Query planning overhead is zero.

24.5 Built-in Lifecycle and Domain Features

25. Boundaries and Debated Topics (Physical Constraints vs. Conscious Architectural Choices)

In database design, every architectural decision serves a specific optimization goal. Certain characteristics that developers coming from traditional SQL environments might initially perceive as “constraints” or “omissions” are, in fact, deliberately engineered core advantages designed to ensure direct index access, deterministic low latency, and maximum I/O throughput.

25.1 Physical and Environmental Boundaries (Out-of-Scope Scenarios)

The following scenarios lie outside the intended operational scope of an embedded, file-based database engine like AmberDB:

25.1.1 High-Concurrency Parallel Write-Heavy Workloads

AmberDB relies on DB_File (Berkeley DB). Write operations enforce a file-level exclusive lock (flock).

25.1.2 Distributed Multi-Node Concurrent Network Writes (Multi-Master Clustering)

AmberDB is optimized for high-speed local filesystem storage. Multiple physical servers writing concurrently to the same database files over shared network storage (e.g., NFS, SMB shares) can encounter lock latency and filesystem cache invalidation delays.

25.2 Debated Topics: Omission or Conscious Performance Advantage?

The following architectural choices might appear restrictive from an ad-hoc SQL mindset, but they are the exact reasons why AmberDB delivers superior throughput and latency:

25.2.1 Full-Table Ad-Hoc Queries on Unindexed Fields: Omission or Performance Guarantee?

25.2.2 Bulk Methods Bypass Undo Journals: Limitation or Maximum I/O Throughput?

25.2.3 Fixed Binary Key Lengths: Limitation or Zero-Copy Slicing Speed?

This documentation is maintained for AmberDB v5.23.0 and aligns with active codebase architecture and developer practices.

This site is open source. Improve this page.