Use suffix array index for search.

10 files changed, 439 insertions, 166 deletions

diff --git a/.gitignore b/.gitignore
index 6eac243..cc4d7ac 100644
--- a/.gitignore
+++ b/.gitignore
@@ -1,4 +1,5 @@
 .jekyll-cache/
 
 **/*.swp
-**/search_index.dat
+**/*.dat
+**/*.bin
diff --git a/_log/arduino-due.md b/_log/arduino-due.md
index f604e89..a0bfe4a 100644
--- a/_log/arduino-due.md
+++ b/_log/arduino-due.md
@@ -1,5 +1,5 @@
 ---
-title: ATSAM3X8E bare-metal programming
+title: Bare-metal ATSAM3X8E
 date: 2024-09-16
 layout: post
 ---
diff --git a/_site/cgi-bin/find.cgi b/_site/cgi-bin/find.cgi
index fc1d4af..5f95e3a 100644
--- a/_site/cgi-bin/find.cgi
+++ b/_site/cgi-bin/find.cgi
@@ -1,86 +1,202 @@
 #!/usr/bin/perl
 
-use Encode qw(decode_utf8);
+use strict;
+use warnings;
+use Storable qw(retrieve);
+use Encode qw(decode_utf8 encode_utf8);
+use URI::Escape qw(uri_unescape);
 use HTML::Escape qw(escape_html);
 
+# --- Configuration ---
+my $max_parallel   = 50;                    # Max parallel search requests
+my $lock_timeout   = 30;                    # Seconds before dropping stale locks
+my $max_results    = 20;                    # Max search results to display
+my $sa_file        = 'sa.bin';              # Suffix Array index
+my $cp_file        = 'corpus.bin';          # Raw text corpus
+my $map_file       = 'file_map.dat';        # File metadata
+my $lock_dir       = '/tmp/search_locks';   # Semaphore directory
+
+# --- Concurrency Control ---
+mkdir $lock_dir, 0777 unless -d $lock_dir;
+my $active_count = 0;
+my $now = time();
+
+opendir(my $dh, $lock_dir);
+while (my $file = readdir($dh)) {
+    next unless $file =~ /\.lock$/;
+    my $path = "$lock_dir/$file";
+    my $mtime = (stat($path))[9] || 0;
+    ($now - $mtime > $lock_timeout) ? unlink($path) : $active_count++;
+}
+closedir($dh);
+
+# Template variables
+my $year = (localtime)[5] + 1900;
 my $search_text = '';
 
-if ($ENV{QUERY_STRING} =~ /^q=([^&]*)/) {
-    $search_text = decode_utf8($1 // "");
-    $search_text =~ s/\P{Print}//g;  # toss any non-printable utf-8 characters
+# Busy check
+if ($active_count >= $max_parallel) {
+    print "Content-Type: text/html\n\n";
+    render_html("<p>Server busy. Please try again in a few seconds.</p>", "", $year);
+    exit;
+}
+
+# Create semaphore lock
+my $lock_file = "$lock_dir/$$.lock";
+open(my $fh_lock, '>', $lock_file);
+
+# --- Query Decoding ---
+if (($ENV{QUERY_STRING} || '') =~ /^q=([^&]*)/) {
+    my $raw_q = $1;
+    $raw_q =~ tr/+/ /;
+    $search_text = uri_unescape($raw_q);
+    $search_text = decode_utf8($search_text // "");
+    $search_text =~ s/\P{Print}//g; 
     $search_text = substr($search_text, 0, 64);
     $search_text =~ s/^\s+|\s+$//g;
 }
 
-my @results;
-
-# Search only index.html files inside the first level of subdirectories
-my $start_dir = '../log'; 
-my @files = glob("$start_dir/*/index.html");
-
-foreach my $path (@files) {
-    # Skip if the path is a symlink or not a file
-    next if -l $path || ! -f $path;
-
-    next unless open(my $fh, '<:utf8', $path);
-    my $html = do { local $/; <$fh> };
-    close($fh);
-  
-    my ($text) = $html =~ m|<main>(.*?)</main>|is;
-    $text =~ s|<[^>]+>| |g;
-    $text =~ s|\s+| |g;
-
-    next unless $text =~ /(.{0,40})(\Q$search_text\E)(.{0,40})/is;
-    my ($before, $actual, $after) = ($1, $2, $3);
-
-    # Trim if we cut into the middle of a sentence
-    $after =~ s/\s\S*$// if length($after) > 25;
-    $before =~ s/^.*?\s// if length($before) > 25;
-
-    if ($before =~ /\S/) { # If before has non-whitespace characters
-        $before = ucfirst($before);
-    } else {
-        $before = "";      # Clear any stray spaces
-        $actual = ucfirst($actual);
-    }
+my $safe_search_text = escape_html($search_text);
 
-    my $safe_before = escape_html($before);
-    my $safe_actual = escape_html($actual);
-    my $safe_after  = escape_html($after);
-    my $snippet = "${safe_before}<b>${safe_actual}</b>${safe_after}...";
+print "Content-Type: text/html\n\n";
 
-    my ($title) = $html =~ m|<title>(.*?)</title>|is;
-    my $safe_title  = escape_html($title);
+if ($search_text eq '') {
+    final_output("<p>Please enter a search term above.</p>");
+}
 
-    $path =~ s|^\.\./||;
+# --- Binary Search Logic ---
+my @results;
+my $query = encode_utf8(lc($search_text));
+my $query_len = length($query);
+
+if (-f $sa_file && -f $cp_file) {
+    open(my $fh_sa, '<', $sa_file) or die $!;
+    open(my $fh_cp, '<', $cp_file) or die $!;
+    binmode($fh_sa);
+    binmode($fh_cp);
+
+    my $file_map = retrieve($map_file);
+    my $total_suffixes = (-s $sa_file) / 4;
+
+    # Find left boundary
+    my ($low, $high) = (0, $total_suffixes - 1);
+    my $first_hit = -1;
+
+    while ($low <= $high) {
+        my $mid = int(($low + $high) / 2);
+        seek($fh_sa, $mid * 4, 0);
+        read($fh_sa, my $bin_off, 4);
+        my $off = unpack("L", $bin_off);
+        seek($fh_cp, $off, 0);
+        read($fh_cp, my $text, $query_len);
+        
+        my $cmp = $text cmp $query;
+        if ($cmp >= 0) {
+            $first_hit = $mid if $cmp == 0;
+            $high = $mid - 1;
+        } else {
+            $low = $mid + 1;
+        }
+    }
 
-    push @results, { 
-        path    => $path,
-        title   => $safe_title, 
-        snippet => $snippet 
-    };
+    # Collect results if found
+    if ($first_hit != -1) {
+        my $last_hit = $first_hit;
+        ($low, $high) = ($first_hit, $total_suffixes - 1);
+        
+        # Find right boundary
+        while ($low <= $high) {
+            my $mid = int(($low + $high) / 2);
+            seek($fh_sa, $mid * 4, 0);
+            read($fh_sa, my $bin_off, 4);
+            my $off = unpack("L", $bin_off);
+            seek($fh_cp, $off, 0);
+            read($fh_cp, my $text, $query_len);
+            
+            if (($text cmp $query) <= 0) {
+                $last_hit = $mid if $text eq $query;
+                $low = $mid + 1;
+            } else {
+                $high = $mid - 1;
+            }
+        }
+
+        my %seen;
+        for my $i ($first_hit .. $last_hit) {
+            seek($fh_sa, $i * 4, 0);
+            read($fh_sa, my $bin_off, 4);
+            my $offset = unpack("L", $bin_off);
+
+            foreach my $m (@$file_map) {
+                if ($offset >= $m->{start} && $offset < $m->{end}) {
+                    if (!$seen{$m->{path}}++) {
+                        # 1. Capture slightly more than 50 chars for trimming
+                        my $snip_start = ($offset - 30 < $m->{start}) ? $m->{start} : $offset - 30;
+                        seek($fh_cp, $snip_start, 0);
+                        read($fh_cp, my $raw_snip, 120);
+                        
+                        my $snippet = decode_utf8($raw_snip, Encode::FB_QUIET) // $raw_snip;
+                        $snippet =~ s/\s+/ /g; # Normalize whitespace
+
+                        # 2. Trim Start: Partial word removal
+                        if ($snip_start > $m->{start}) {
+                            $snippet =~ s/^[^\s]*\s//; 
+                        }
+
+                        # 3. Trim End: Length limit and partial word removal
+                        my $has_more = 0;
+                        if (length($snippet) > 50) {
+                            $snippet = substr($snippet, 0, 50);
+                            $has_more = 1 if $snippet =~ s/\s+[^\s]*$//;
+                        }
+
+                        # 4. Cleanup & Capitalize
+                        $snippet = ucfirst($snippet);
+                        my $clean_path = $m->{path};
+                        $clean_path =~ s|^\.\./_site/||;
+
+                        # 5. Build Final Snippet
+                        my $display_snippet = escape_html($snippet) . ($has_more ? "..." : "");
+
+                        push @results, {
+                            path    => $clean_path,
+                            title   => (split('/', $m->{path}))[-2],
+                            snippet => $display_snippet
+                        };
+                    }
+                    last;
+                }
+            }
+            last if scalar @results >= $max_results;
+        }
+    }
+    close($fh_sa);
+    close($fh_cp);
 }
 
-print "Content-Type: text/html\n\n";
-
-my $list;
-if ($search_text eq '') {
-    $list = "<p>Please enter a search term above.</p>";
-} elsif (@results == 0) {
-    $list = "<p>No results found for \"<b>$search_text</b>\".</p>";
+# --- Formatting & Output ---
+my $list_html = "";
+if (@results == 0) {
+    $list_html = "<p>No results found for \"<b>$safe_search_text</b>\".</p>";
 } else {
-    $list = "<ul>";
-    foreach my $res (@results) {
-        my $url = $res->{path};
-        $list .= "<li><a href=\"/$url\">$res->{title}</a><br><small>$res->{snippet}</small></li>";
-    }
-    $list .= "</ul>";
+    $list_html = "<ul>" . join('', map { 
+        "<li><a href=\"/$_->{path}\">$_->{title}</a><br><small>$_->{snippet}</small></li>" 
+    } @results) . "</ul>";
 }
 
-my $safe_search_text = escape_html($search_text);
-my $year = (localtime)[5] + 1900;
+final_output($list_html);
+
+# --- Helpers ---
+sub final_output {
+    my ($content) = @_;
+    render_html($content, $safe_search_text, $year);
+    if ($fh_lock) { close($fh_lock); unlink($lock_file); }
+    exit;
+}
 
-print <<"HTML";
+sub render_html {
+    my ($content, $q_val, $yr) = @_;
+    print <<"HTML";
 <!DOCTYPE html>
 <html lang="en-us">
 <head>
@@ -105,19 +221,21 @@ print <<"HTML";
         <div class="container">
             <h2>Search</h2>
             <form action="" method="GET">
-                <input id="search-box" type="text" name="q" value="$safe_search_text">
+                <input id="search-box" type="text" name="q" value="$q_val">
                 <input id="search-btn" type="submit" value="Search">
             </form>
-            $list
+            $content
         </div>
     </main>
     <div class="footer">
         <div class="container">
             <div class="twelve columns right container-2">
-                <p id="footer-text">&copy; ASCIIMX - $year</p>
+                <p id="footer-text">&copy; ASCIIMX - $yr</p>
             </div>
         </div>
     </div>
 </body>
 </html>
 HTML
+}
+
diff --git a/_site/feed.xml b/_site/feed.xml
index dfe24b2..9ccb174 100644
--- a/_site/feed.xml
+++ b/_site/feed.xml
@@ -1 +1 @@
-<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="/feed.xml" rel="self" type="application/atom+xml" /><link href="/" rel="alternate" type="text/html" /><updated>2026-01-01T16:23:09+08:00</updated><id>/feed.xml</id><title type="html">ASCIIMX | Log</title><author><name>W. D. Sadeep Madurange</name></author><entry><title type="html">Matrix Rain: 2025 refactor</title><link href="/log/matrix-digital-rain/" rel="alternate" type="text/html" title="Matrix Rain: 2025 refactor" /><published>2025-12-21T00:00:00+08:00</published><updated>2025-12-21T00:00:00+08:00</updated><id>/log/matrix-digital-rain</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[The 2022 version worked but had some loose ends. Unicode support was inflexible–couldn’t mix ASCII with Katakana; Phosphor decay was stored in a separate array when it should’ve been packed with RGB; Code was harder to read than it needed to be.]]></summary></entry><entry><title type="html">Fingerprint door lock (LP)</title><link href="/log/fpm-door-lock-lp/" rel="alternate" type="text/html" title="Fingerprint door lock (LP)" /><published>2025-08-18T00:00:00+08:00</published><updated>2025-08-18T00:00:00+08:00</updated><id>/log/fpm-door-lock-lp</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Second iteration of the RF door lock. Old version worked but drew too much quiescent current. Sensor and servo pulled 13.8mA and 4.6mA idle. Linear regulators were a disaster. Battery didn’t last 24 hours.]]></summary></entry><entry><title type="html">High-side MOSFET switching</title><link href="/log/mosfet-switches/" rel="alternate" type="text/html" title="High-side MOSFET switching" /><published>2025-06-22T00:00:00+08:00</published><updated>2025-06-22T00:00:00+08:00</updated><id>/log/mosfet-switches</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Needed low-power switching for the fingerprint door lock. Servo and FPM draw high quiescent current–had to cut power electronically during sleep. MOSFETs can do this.]]></summary></entry><entry><title type="html">ATmega328P at 3.3V and 5V</title><link href="/log/arduino-uno/" rel="alternate" type="text/html" title="ATmega328P at 3.3V and 5V" /><published>2025-06-10T00:00:00+08:00</published><updated>2025-06-10T00:00:00+08:00</updated><id>/log/arduino-uno</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Quick reference for wiring ATmega328P ICs at 5V and 3.3V. 5V uses 16MHz crystal, 3.3V uses 8MHz.]]></summary></entry><entry><title type="html">Fingerprint door lock (RF)</title><link href="/log/fpm-door-lock-rf/" rel="alternate" type="text/html" title="Fingerprint door lock (RF)" /><published>2025-06-05T00:00:00+08:00</published><updated>2025-06-05T00:00:00+08:00</updated><id>/log/fpm-door-lock-rf</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Wanted to unlock door with fingerprint, wirelessly to avoid drilling.]]></summary></entry><entry><title type="html">Bumblebee: browser automation</title><link href="/log/bumblebee/" rel="alternate" type="text/html" title="Bumblebee: browser automation" /><published>2025-04-02T00:00:00+08:00</published><updated>2025-04-02T00:00:00+08:00</updated><id>/log/bumblebee</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Built with Andy Zhang for an employer. Tool to automate web scraping script generation.]]></summary></entry><entry><title type="html">ATSAM3X8E bare-metal programming</title><link href="/log/arduino-due/" rel="alternate" type="text/html" title="ATSAM3X8E bare-metal programming" /><published>2024-09-16T00:00:00+08:00</published><updated>2024-09-16T00:00:00+08:00</updated><id>/log/arduino-due</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Notes on programming bare-metal ATSAM3X8E chips (Arduino Due) using Serial Wire Debug (SwD) protocol.]]></summary></entry><entry><title type="html">Etlas: e-paper dashboard</title><link href="/log/etlas/" rel="alternate" type="text/html" title="Etlas: e-paper dashboard" /><published>2024-09-05T00:00:00+08:00</published><updated>2024-09-05T00:00:00+08:00</updated><id>/log/etlas</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Repurposed e-reader prototype into something for regular use. News, stocks, weather dashboard. ESP32 NodeMCU D1 + 7.5” Waveshare e-paper + DHT22 sensor.]]></summary></entry><entry><title type="html">ESP32 e-reader prototype</title><link href="/log/e-reader/" rel="alternate" type="text/html" title="ESP32 e-reader prototype" /><published>2023-10-24T00:00:00+08:00</published><updated>2023-10-24T00:00:00+08:00</updated><id>/log/e-reader</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[First project with e-paper displays and ESP32.]]></summary></entry><entry><title type="html">Neo4j shortest path optimization</title><link href="/log/neo4j-a-star-search/" rel="alternate" type="text/html" title="Neo4j shortest path optimization" /><published>2018-03-06T00:00:00+08:00</published><updated>2018-03-06T00:00:00+08:00</updated><id>/log/neo4j-a-star-search</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Replaced Dijkstra’s search for vessel route tracking in Neo4J.]]></summary></entry></feed>
\ No newline at end of file
+<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="/feed.xml" rel="self" type="application/atom+xml" /><link href="/" rel="alternate" type="text/html" /><updated>2026-01-02T19:06:26+08:00</updated><id>/feed.xml</id><title type="html">ASCIIMX | Log</title><author><name>W. D. Sadeep Madurange</name></author><entry><title type="html">Matrix Rain: 2025 refactor</title><link href="/log/matrix-digital-rain/" rel="alternate" type="text/html" title="Matrix Rain: 2025 refactor" /><published>2025-12-21T00:00:00+08:00</published><updated>2025-12-21T00:00:00+08:00</updated><id>/log/matrix-digital-rain</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[The 2022 version worked but had some loose ends. Unicode support was inflexible–couldn’t mix ASCII with Katakana; Phosphor decay was stored in a separate array when it should’ve been packed with RGB; Code was harder to read than it needed to be.]]></summary></entry><entry><title type="html">Fingerprint door lock (LP)</title><link href="/log/fpm-door-lock-lp/" rel="alternate" type="text/html" title="Fingerprint door lock (LP)" /><published>2025-08-18T00:00:00+08:00</published><updated>2025-08-18T00:00:00+08:00</updated><id>/log/fpm-door-lock-lp</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Second iteration of the RF door lock. Old version worked but drew too much quiescent current. Sensor and servo pulled 13.8mA and 4.6mA idle. Linear regulators were a disaster. Battery didn’t last 24 hours.]]></summary></entry><entry><title type="html">High-side MOSFET switching</title><link href="/log/mosfet-switches/" rel="alternate" type="text/html" title="High-side MOSFET switching" /><published>2025-06-22T00:00:00+08:00</published><updated>2025-06-22T00:00:00+08:00</updated><id>/log/mosfet-switches</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Needed low-power switching for the fingerprint door lock. Servo and FPM draw high quiescent current–had to cut power electronically during sleep. MOSFETs can do this.]]></summary></entry><entry><title type="html">ATmega328P at 3.3V and 5V</title><link href="/log/arduino-uno/" rel="alternate" type="text/html" title="ATmega328P at 3.3V and 5V" /><published>2025-06-10T00:00:00+08:00</published><updated>2025-06-10T00:00:00+08:00</updated><id>/log/arduino-uno</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Quick reference for wiring ATmega328P ICs at 5V and 3.3V. 5V uses 16MHz crystal, 3.3V uses 8MHz.]]></summary></entry><entry><title type="html">Fingerprint door lock (RF)</title><link href="/log/fpm-door-lock-rf/" rel="alternate" type="text/html" title="Fingerprint door lock (RF)" /><published>2025-06-05T00:00:00+08:00</published><updated>2025-06-05T00:00:00+08:00</updated><id>/log/fpm-door-lock-rf</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Wanted to unlock door with fingerprint, wirelessly to avoid drilling.]]></summary></entry><entry><title type="html">Bumblebee: browser automation</title><link href="/log/bumblebee/" rel="alternate" type="text/html" title="Bumblebee: browser automation" /><published>2025-04-02T00:00:00+08:00</published><updated>2025-04-02T00:00:00+08:00</updated><id>/log/bumblebee</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Built with Andy Zhang for an employer. Tool to automate web scraping script generation.]]></summary></entry><entry><title type="html">Bare-metal ATSAM3X8E</title><link href="/log/arduino-due/" rel="alternate" type="text/html" title="Bare-metal ATSAM3X8E" /><published>2024-09-16T00:00:00+08:00</published><updated>2024-09-16T00:00:00+08:00</updated><id>/log/arduino-due</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Notes on programming bare-metal ATSAM3X8E chips (Arduino Due) using Serial Wire Debug (SwD) protocol.]]></summary></entry><entry><title type="html">Etlas: e-paper dashboard</title><link href="/log/etlas/" rel="alternate" type="text/html" title="Etlas: e-paper dashboard" /><published>2024-09-05T00:00:00+08:00</published><updated>2024-09-05T00:00:00+08:00</updated><id>/log/etlas</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Repurposed e-reader prototype into something for regular use. News, stocks, weather dashboard. ESP32 NodeMCU D1 + 7.5” Waveshare e-paper + DHT22 sensor.]]></summary></entry><entry><title type="html">ESP32 e-reader prototype</title><link href="/log/e-reader/" rel="alternate" type="text/html" title="ESP32 e-reader prototype" /><published>2023-10-24T00:00:00+08:00</published><updated>2023-10-24T00:00:00+08:00</updated><id>/log/e-reader</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[First project with e-paper displays and ESP32.]]></summary></entry><entry><title type="html">Neo4j shortest path optimization</title><link href="/log/neo4j-a-star-search/" rel="alternate" type="text/html" title="Neo4j shortest path optimization" /><published>2018-03-06T00:00:00+08:00</published><updated>2018-03-06T00:00:00+08:00</updated><id>/log/neo4j-a-star-search</id><author><name>W. D. Sadeep Madurange</name></author><summary type="html"><![CDATA[Replaced Dijkstra’s search for vessel route tracking in Neo4J.]]></summary></entry></feed>
diff --git a/_site/index.html b/_site/index.html
index 257c38d..90cbbfd 100644
--- a/_site/index.html
+++ b/_site/index.html
@@ -139,7 +139,7 @@
   
       <tr>
         <td class="posts-td posts-td-link">
-          <a href="/log/arduino-due/" class="link-decor-none">ATSAM3X8E bare-metal programming</a>
+          <a href="/log/arduino-due/" class="link-decor-none">Bare-metal ATSAM3X8E</a>
         </td>
         <td class="posts-td posts-td-time">
           <span class="post-meta">
diff --git a/_site/log/arduino-due/index.html b/_site/log/arduino-due/index.html
index d8178c4..7bc8a0b 100644
--- a/_site/log/arduino-due/index.html
+++ b/_site/log/arduino-due/index.html
@@ -3,7 +3,7 @@
     <head>
   <meta charset="utf-8">
   <meta name="viewport" content="width=device-width, initial-scale=1">
-  <title>ATSAM3X8E bare-metal programming</title>
+  <title>Bare-metal ATSAM3X8E</title>
   <link rel="stylesheet" href="/assets/css/main.css">
   <link rel="stylesheet" href="/assets/css/skeleton.css">
 </head>
@@ -40,7 +40,7 @@
     <main>
       <div class="container">
         <div class="container-2">
-          <h2 class="center" id="title">ATSAM3X8E BARE-METAL PROGRAMMING</h2>
+          <h2 class="center" id="title">BARE-METAL ATSAM3X8E</h2>
           <h5 class="center">16 SEPTEMBER 2024</h5>
           <br>
           <div class="twocol justify"><p>Notes on programming bare-metal ATSAM3X8E chips (Arduino Due) using Serial Wire
diff --git a/_site/log/index.html b/_site/log/index.html
index 93d2573..32fbf8d 100644
--- a/_site/log/index.html
+++ b/_site/log/index.html
@@ -129,7 +129,7 @@
   
       <tr>
         <td class="posts-td posts-td-link">
-          <a href="/log/arduino-due/" class="link-decor-none">ATSAM3X8E bare-metal programming</a>
+          <a href="/log/arduino-due/" class="link-decor-none">Bare-metal ATSAM3X8E</a>
         </td>
         <td class="posts-td posts-td-time">
           <span class="post-meta">
diff --git a/_site/posts.xml b/_site/posts.xml
index 5677165..d327a7b 100644
--- a/_site/posts.xml
+++ b/_site/posts.xml
@@ -1 +1 @@
-<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="/posts.xml" rel="self" type="application/atom+xml" /><link href="/" rel="alternate" type="text/html" /><updated>2026-01-01T16:23:09+08:00</updated><id>/posts.xml</id><title type="html">ASCIIMX</title><author><name>W. D. Sadeep Madurange</name></author></feed>
\ No newline at end of file
+<?xml version="1.0" encoding="utf-8"?><feed xmlns="http://www.w3.org/2005/Atom" ><generator uri="https://jekyllrb.com/" version="4.4.1">Jekyll</generator><link href="/posts.xml" rel="self" type="application/atom+xml" /><link href="/" rel="alternate" type="text/html" /><updated>2026-01-02T19:06:26+08:00</updated><id>/posts.xml</id><title type="html">ASCIIMX</title><author><name>W. D. Sadeep Madurange</name></author></feed>
diff --git a/cgi-bin/find.cgi b/cgi-bin/find.cgi
index 3e5c8a2..ab066dd 100644
--- a/cgi-bin/find.cgi
+++ b/cgi-bin/find.cgi
@@ -3,16 +3,18 @@
 use strict;
 use warnings;
 use Storable qw(retrieve);
-use Encode qw(decode_utf8);
+use Encode qw(decode_utf8 encode_utf8);
+use URI::Escape qw(uri_unescape);
 use HTML::Escape qw(escape_html);
 
 # Configuration
-my $max_parallel   = 50;                    # max no. of parallel searches         
-my $lock_timeout   = 30;                    # drop stale locks after this many seconds
-my $max_results    = 20;                    # max search results
-my $min_query_len  = 3;                     # min query length to avoid matching 'a', 'e'
-my $index_file     = 'search_index.dat';    # index file
-my $lock_dir       = '/tmp/search_locks';   # lock file directory
+my $max_parallel   = 50;                    # Max parallel search requests
+my $lock_timeout   = 30;                    # Seconds before dropping stale locks
+my $max_results    = 20;                    # Max search results to display
+my $sa_file        = 'sa.bin';              # Suffix Array index
+my $cp_file        = 'corpus.bin';          # Raw text corpus
+my $map_file       = 'file_map.dat';        # File metadata
+my $lock_dir       = '/tmp/search_locks';   # Semaphore directory
 
 # Concurrency control
 mkdir $lock_dir, 0777 unless -d $lock_dir;
@@ -24,78 +26,161 @@ while (my $file = readdir($dh)) {
     next unless $file =~ /\.lock$/;
     my $path = "$lock_dir/$file";
     my $mtime = (stat($path))[9] || 0;
-    ( $now - $mtime > $lock_timeout ) ? unlink($path) : $active_count++;
+    ($now - $mtime > $lock_timeout) ? unlink($path) : $active_count++;
 }
 closedir($dh);
 
-# Too many search requests
+# Template variables
+my $year = (localtime)[5] + 1900;
+my $search_text = '';
+
+# Busy check
 if ($active_count >= $max_parallel) {
     print "Content-Type: text/html\n\n";
-    render_html("<p>Server busy. Please try again in a few seconds.</p>", "", (localtime)[5]+1900);
+    render_html("<p>Server busy. Please try again in a few seconds.</p>", "", $year);
     exit;
 }
 
+# Create semaphore lock
 my $lock_file = "$lock_dir/$$.lock";
 open(my $fh_lock, '>', $lock_file);
 
-# Decode search text as utf-8, toss non-printable chars, trim
-my $search_text = '';
+# Query decoding
 if (($ENV{QUERY_STRING} || '') =~ /^q=([^&]*)/) {
-    $search_text = decode_utf8($1 // "");
+    my $raw_q = $1;
+    $raw_q =~ tr/+/ /;
+    $search_text = uri_unescape($raw_q);
+    $search_text = decode_utf8($search_text // "");
     $search_text =~ s/\P{Print}//g; 
     $search_text = substr($search_text, 0, 64);
     $search_text =~ s/^\s+|\s+$//g;
 }
 
-# Pre-prepare common template variables
 my $safe_search_text = escape_html($search_text);
-my $year = (localtime)[5] + 1900;
 
 print "Content-Type: text/html\n\n";
 
-# Input validation
 if ($search_text eq '') {
     final_output("<p>Please enter a search term above.</p>");
 }
 
-if (length($search_text) < $min_query_len) {
-    final_output("<p>Search term is too short. Please enter at least $min_query_len characters.</p>");
-}
-
-if (!-f $index_file) {
-    final_output("<p>Search temporarily unavailable.</p>");
-}
-
-my $index = retrieve($index_file);
+# Binary search 
 my @results;
-my $found = 0;
-
-foreach my $url (sort keys %$index) {
-    last if $found >= $max_results;
-    my $data = $index->{$url};
-
-    # Grab 80 char snippet to chop at a word boundary later
-    next unless $data->{c} =~ /(.{0,40})(\Q$search_text\E)(.{0,40})/is;
-    my ($before, $actual, $after) = ($1, $2, $3);
-    $found++;
-
-    # Chop at 25 or word boundary
-    $after =~ s/\s\S*$// if length($after) > 25;
-    $before =~ s/^.*?\s// if length($before) > 25;
-
-    $before = ($before =~ /\S/) ? ucfirst($before) : "";
-    $actual = ($before eq "") ? ucfirst($actual) : $actual;
-
-    my $snippet = escape_html($before) . "<b>" . escape_html($actual) . "</b>" . escape_html($after) . "...";
-
-    push @results, { 
-        path    => $url,
-        title   => escape_html($data->{t}), 
-        snippet => $snippet 
-    };
+my $query = encode_utf8(lc($search_text));
+my $query_len = length($query);
+
+if (-f $sa_file && -f $cp_file) {
+    open(my $fh_sa, '<', $sa_file) or die $!;
+    open(my $fh_cp, '<', $cp_file) or die $!;
+    binmode($fh_sa);
+    binmode($fh_cp);
+
+    my $file_map = retrieve($map_file);
+    my $total_suffixes = (-s $sa_file) / 4;
+
+    # Find left boundary
+    my ($low, $high) = (0, $total_suffixes - 1);
+    my $first_hit = -1;
+
+    while ($low <= $high) {
+        my $mid = int(($low + $high) / 2);
+        seek($fh_sa, $mid * 4, 0);
+        read($fh_sa, my $bin_off, 4);
+        my $off = unpack("L", $bin_off);
+        seek($fh_cp, $off, 0);
+        read($fh_cp, my $text, $query_len);
+        
+        my $cmp = $text cmp $query;
+        if ($cmp >= 0) {
+            $first_hit = $mid if $cmp == 0;
+            $high = $mid - 1;
+        } else {
+            $low = $mid + 1;
+        }
+    }
+
+    # Collect results if found
+    if ($first_hit != -1) {
+        my $last_hit = $first_hit;
+        ($low, $high) = ($first_hit, $total_suffixes - 1);
+        
+        # Find right boundary
+        while ($low <= $high) {
+            my $mid = int(($low + $high) / 2);
+            seek($fh_sa, $mid * 4, 0);
+            read($fh_sa, my $bin_off, 4);
+            my $off = unpack("L", $bin_off);
+            seek($fh_cp, $off, 0);
+            read($fh_cp, my $text, $query_len);
+            
+            if (($text cmp $query) <= 0) {
+                $last_hit = $mid if $text eq $query;
+                $low = $mid + 1;
+            } else {
+                $high = $mid - 1;
+            }
+        }
+
+        my %seen;
+        for my $i ($first_hit .. $last_hit) {
+            seek($fh_sa, $i * 4, 0);
+            read($fh_sa, my $bin_off, 4);
+            my $offset = unpack("L", $bin_off);
+
+            foreach my $m (@$file_map) {
+                if ($offset >= $m->{start} && $offset < $m->{end}) {
+                    if (!$seen{$m->{path}}++) {
+                        # Capture more than 50 chars for trimming
+                        my $snip_start = ($offset - 30 < $m->{start}) ? $m->{start} : $offset - 30;
+                        my $max_len = $m->{end} - $snip_start;
+                        my $read_len = ($max_len > 120) ? 120 : $max_len;
+                        seek($fh_cp, $snip_start, 0);
+                        read($fh_cp, my $raw_snip, $read_len);
+                        
+                        my $snippet = decode_utf8($raw_snip, Encode::FB_QUIET) // $raw_snip;
+                        $snippet =~ s/\s+/ /g; # Normalize whitespace
+
+                        # Trim start: Partial word removal
+                        if ($snip_start > $m->{start}) {
+                            $snippet =~ s/^[^\s]*\s//; 
+                        }
+
+                        # Trim end: Length limit and partial word removal
+                        my $has_more = 0;
+                        if (length($snippet) > 50) {
+                            $snippet = substr($snippet, 0, 50);
+                            $has_more = 1 if $snippet =~ s/\s+[^\s]*$//;
+                        }
+                        elsif ($snip_start + $read_len < $m->{end}) {
+                            # This check handles snippets that are naturally short but 
+                            # there's still more text in the article we didn't read
+                            $has_more = 1; 
+                        }
+
+                        # Cleanup & capitalize
+                        $snippet = ucfirst($snippet);
+                        $snippet = escape_html($snippet) . ($has_more ? "..." : "");
+
+                        my $clean_path = $m->{path};
+                        $clean_path =~ s|^\.\./_site/||;
+
+                        push @results, {
+                            path    => $clean_path,
+                            title   => $m->{title},,
+                            snippet => $snippet
+                        };
+                    }
+                    last;
+                }
+            }
+            last if scalar @results >= $max_results;
+        }
+    }
+    close($fh_sa);
+    close($fh_cp);
 }
 
-# Format results list
+# --- Formatting & Output ---
 my $list_html = "";
 if (@results == 0) {
     $list_html = "<p>No results found for \"<b>$safe_search_text</b>\".</p>";
@@ -107,12 +192,11 @@ if (@results == 0) {
 
 final_output($list_html);
 
-# Helper to ensure layout is always preserved
+# --- Helpers ---
 sub final_output {
     my ($content) = @_;
     render_html($content, $safe_search_text, $year);
-    close($fh_lock) if $fh_lock;
-    unlink($lock_file) if -f $lock_file;
+    if ($fh_lock) { close($fh_lock); unlink($lock_file); }
     exit;
 }
 
diff --git a/cgi-bin/indexer.pl b/cgi-bin/indexer.pl
index 38a918e..69f6838 100644
--- a/cgi-bin/indexer.pl
+++ b/cgi-bin/indexer.pl
@@ -2,45 +2,115 @@
 
 use strict;
 use warnings;
-use Storable qw(nstore);
+use File::Find;
+use Storable qw(store);
+use Encode qw(encode_utf8);
 use HTML::Entities qw(decode_entities);
+use Time::HiRes qw(gettimeofday tv_interval);
 
-# --- Configuration ---
-my $built_site_dir = '../_site/log'; 
-my $output_file    = '../_site/cgi-bin/search_index.dat';
-my %index;
-
-print "Building search index from $built_site_dir...\n";
-
-foreach my $path (glob("$built_site_dir/*/index.html")) {
-    next unless open(my $fh, '<:utf8', $path);
-    my $html = do { local $/; <$fh> };
-    close($fh);
-
-    # Extract Title and Main Content
-    my ($title) = $html =~ m|<title>(.*?)</title>|is;
-    my ($main)  = $html =~ m|<main>(.*?)</main>|is;
-    $main //= '';
-
-    # Strip HTML and clean prose
-    $main =~ s|<pre[^>]*>.*?</pre>| |gs;
-    $main =~ s|<code[^>]*>.*?</code>| |gs;
-    $main =~ s|<[^>]+>| |g;
-    $main = decode_entities($main);
-    $main =~ s|\s+| |g;
-    $main =~ s/^\s+|\s+$//g;
-
-    # Normalize path
-    my $url = $path;
-    $url =~ s|^\.\./_site/||; # Remove local build directory
-    $url =~ s|^\.\./||;       # Remove any leading dots
-    $url =~ s|^/+||;          # Remove leading slashes
-
-    $index{$url} = {
-        t => $title || "Untitled",
-        c => $main
-    };
+my $dir         = '../_site/log';
+my $cgi_dir     = '../_site/cgi-bin/';
+my $corpus_file = "${cgi_dir}corpus.bin";
+my $sa_file     = "${cgi_dir}sa.bin";
+my $map_file    = "${cgi_dir}file_map.dat";
+
+my %excluded_files = (
+    'index.html'  => 1, # /log/index.html
+);
+
+# Start timing
+my $t0 = [gettimeofday];
+
+my $corpus = "";
+my @file_map;
+
+print "Building corpus...\n";
+
+find({
+    wanted => sub {
+        # Only index index.html files
+        return unless -f $_ && $_ eq 'index.html';
+
+        my $rel_path = $File::Find::name;
+        $rel_path =~ s|^\Q$dir\E/?||; 
+        return if $excluded_files{$rel_path};
+        
+        if (open my $fh, '<:encoding(UTF-8)', $_) {
+            my $content = do { local $/; <$fh> };
+            close $fh;
+
+            my ($title) = $content =~ m|<title>(.*?)</title>|is;
+            $title //= (split('/', $File::Find::name))[-2]; # Fallback to folder name
+            $title =~ s/^\s+|\s+$//g;
+
+            # Extract content from <main> or use whole file
+            my ($text) = $content =~ m|<main>(.*?)</main>|is;
+            $text //= $content; 
+
+            # Strip tags and normalize whitespace
+            $text =~ s|<pre[^>]*>.*?</pre>| |gs;
+            $text =~ s|<code[^>]*>.*?</code>| |gs;
+            $text =~ s|<[^>]+>| |g;
+            $text = decode_entities($text);
+            $text =~ s|\s+| |g;
+            $text =~ s/^\s+|\s+$//g;
+
+            # CRITICAL: Convert to lowercase and then to raw bytes
+            # This ensures length() and substr() work on byte offsets for seek()
+            my $raw_entry = encode_utf8(lc($text) . "\0");
+
+            my $start = length($corpus);
+            $corpus .= $raw_entry;
+            
+            push @file_map, { 
+                start => $start, 
+                end   => length($corpus), 
+                title => $title, 
+                path  => $File::Find::name 
+            };
+        }
+    },
+    no_chdir => 0,
+}, $dir);
+
+print "Sorting suffixes...\n";
+
+# Initialize the array of indices
+my @sa = 0 .. (length($corpus) - 1);
+
+# Use a block that forces byte-level comparison
+{
+    use bytes; 
+    @sa = sort { 
+        # First 64 bytes check (fast path)
+        (substr($corpus, $a, 64) cmp substr($corpus, $b, 64)) || 
+        # Full string fallback (required for correctness)
+        (substr($corpus, $a) cmp substr($corpus, $b))
+    } @sa;
 }
 
-nstore(\%index, $output_file);
-printf("Index complete: %d files (%.2f KB)\n", scalar(keys %index), (-s $output_file) / 1024);
+print "Writing index files to disk...\n";
+
+open my $cfh, '>', $corpus_file or die "Cannot write $corpus_file: $!";
+binmode($cfh); # Raw byte mode
+print $cfh $corpus;
+close $cfh;
+
+open my $sfh, '>', $sa_file or die "Cannot write $sa_file: $!";
+binmode($sfh);
+# Pack as 32-bit unsigned integers (standard 'L')
+print $sfh pack("L*", @sa);
+close $sfh;
+
+store \@file_map, $map_file;
+
+my $elapsed = tv_interval($t0);
+my $c_size = -s $corpus_file;
+my $s_size = -s $sa_file;
+
+printf "\nIndexing Complete!\n";
+printf "Total Time:      %.4f seconds\n", $elapsed;
+printf "Corpus Size:     %.2f KB\n", $c_size / 1024;
+printf "Suffix Array:    %.2f KB\n", $s_size / 1024;
+printf "Files Processed: %d\n", scalar(@file_map);
+