slides review

Author: fglock

dev/presentations/German_Perl_Raku_Workshop_2026/Makefile

@@ -1,4 +1,4 @@
-.PHONY: help serve open clean slides part1 part2
+.PHONY: help serve open clean slides part1 part2 part3
 
 # Which slides file to present (default: original combined deck)
 SLIDES ?= slides.md
@@ -10,6 +10,7 @@ help:
 	@echo "  make slides   - Present original combined deck (slides.md)"
 	@echo "  make part1    - Present Part 1: Introduction & Demo (20 min)"
 	@echo "  make part2    - Present Part 2: Technical Deep-Dive (40 min)"
+	@echo "  make part3    - Present Part 3: Integration & Future"
 	@echo "  make serve    - Present with custom file: make serve SLIDES=myfile.md"
 	@echo "  make open     - Just open browser (server must be running)"
 	@echo "  make stats    - Collect statistics and benchmarks for slide numbers"
@@ -27,6 +28,9 @@ part1:
 part2:
 	@$(MAKE) serve SLIDES=slides-part2-technical.md
 
+part3:
+	@$(MAKE) serve SLIDES=slides-part3-integration.md
+
 # Start web server and open browser (use SLIDES variable to choose deck)
 serve:
 	@echo "Presenting: $(SLIDES)"

dev/presentations/German_Perl_Raku_Workshop_2026/slides-part1-intro.md

@@ -107,9 +107,9 @@ Each prior attempt informed this implementation. JPL showed embedding was possib
 ## By the Numbers
 
 - <span class="metric">~200,000 tests</span> in the suite
-- <span class="metric">392 Java source files</span>
-- <span class="metric">5,741 commits</span> since June 2024
-- <span class="metric">2x faster</span> than Perl 5 on loop benchmarks
+- <span class="metric">~400 Java source files</span>
+- <span class="metric">~6,000 commits</span> since June 2024
+- <span class="metric">JIT-optimized</span> hot paths via HotSpot
 
 No formal Perl spec exists — the test suite **is** the specification.
 
@@ -128,22 +128,6 @@ DBI, HTTP::Tiny, JSON, YAML, Text::CSV, Digest::MD5, MIME::Base64…
 
 ---
 
-## Real-World Validation: Image::ExifTool
-
-The most widely-used Perl photo metadata library — running on PerlOnJava **unmodified**.
-
-- <span class="metric">239 Perl source files</span> · <span class="metric">296,000 lines of code</span>
-- Largest modules exceed **10,000 lines** (Nikon.pm, Sony.pm, Canon.pm)
-- Subroutines over **1,000 lines** (SetNewValue, WriteInfo, ExtractInfo)
-- <span class="metric">600 tests</span> in 113 files — **all pass**
-
-**Why this matters:** Large methods exceed the JVM's 64KB bytecode limit — the compiler automatically falls back to the Internal VM. ExifTool proves PerlOnJava handles production-scale Perl reliably, not just toy examples.
-
-Note:
-Image::ExifTool 13.44 by Phil Harvey. Modules like TagLookup.pm (13,840 LOC), Nikon.pm (12,843 LOC), and Writer.pl (5,849 LOC) stress-test every part of the compilation pipeline. The automatic dual-backend fallback is transparent — ExifTool doesn't know which backend runs each method.
-
----
-
 ## Getting Started
 
 ```bash
@@ -180,13 +164,16 @@ Add Maven dependencies with: `./Configure.pl --search mysql-connector-java`. Pur
 ```perl
 use DBI;
 my $dbh = DBI->connect(
-    "dbi:SQLite:dbname=test.db", "", "",
+    "jdbc:h2:mem:testdb;DB_CLOSE_DELAY=-1",
+    "sa", "",
     { RaiseError => 1 }
 );
-my $sth = $dbh->prepare("SELECT * FROM users");
-$sth->execute();
+$dbh->do("INSERT INTO users (name, age) VALUES (?, ?)",
+    undef, "Alice", 30);
+my $sth = $dbh->prepare("SELECT * FROM users WHERE age > ?");
+$sth->execute(20);
 while (my $row = $sth->fetchrow_hashref) {
-    say "User: $row->{name}";
+    say "$row->{name}, age $row->{age}";
 }
 ```
 
@@ -195,6 +182,22 @@ Supports PostgreSQL, MySQL, Oracle, SQLite, H2 — any JDBC driver.
 
 ---
 
+## Live Demo: Image::ExifTool
+
+The most widely-used Perl photo metadata library — running on PerlOnJava **unmodified**.
+
+- <span class="metric">239 Perl source files</span> · <span class="metric">296,000 lines of code</span>
+- Largest modules exceed **10,000 lines** (Nikon.pm, Sony.pm, Canon.pm)
+- Subroutines over **1,000 lines** (SetNewValue, WriteInfo, ExtractInfo)
+- <span class="metric">600 tests</span> in 113 files — **all pass**
+
+**Why this matters:** Large methods exceed the JVM's 64KB bytecode limit — the compiler automatically falls back to the Internal VM.
+
+Note:
+Image::ExifTool 13.44 by Phil Harvey. Modules like TagLookup.pm (13,840 LOC), Nikon.pm (12,843 LOC), and Writer.pl (5,849 LOC) stress-test every part of the compilation pipeline. The automatic dual-backend fallback is transparent — ExifTool doesn't know which backend runs each method.
+
+---
+
 ## How It Works: AST
 
 **Abstract Syntax Tree** — tree representation of source code
@@ -292,14 +295,24 @@ Register-based, ~300 opcodes — much more compact.
 
 ---
 
+## Key Takeaways
+
+- **One JAR, zero dependencies** — drop-in Perl runtime for the JVM
+- **Run existing Perl unchanged** — 200,000+ tests, ExifTool passes unmodified
+- **Full JVM integration** — JDBC, JSR-223, Docker/Kubernetes ready
+- **Two backends, one runtime** — performance where it matters, flexibility where it's needed
+
+---
+
 ## What's Next: Part 2
 
-**Technical deep-dive (40 minutes):**
+**Technical deep-dive (40 minutes)**
 
 - The compilation pipeline in detail
 - Why two backends — and when each one wins
 - JVM optimization techniques applied to Perl
 - Implementing complex Perl semantics on the JVM
-- Integration and future plans
+
+**Integration and future plans**
 
 ---

dev/presentations/German_Perl_Raku_Workshop_2026/slides-part2-technical.md

@@ -577,148 +577,24 @@ Also: globalIORefs → IO, globalFormatRefs → FORMAT. Slot access: *foo{CODE}
 
 ---
 
-## Module Loading
+## Key Takeaways
 
-`require` converts `Module::Name` → `Module/Name.pm`, searches `@INC`, caches in `%INC`.
+- **Dual backend** — JVM for performance, Internal VM for flexibility
+- **Shared runtime** — both backends use the same data structures
+- **JVM-friendly design** — fast/slow path splits enable JIT inlining
+- **Perl's complexity mapped to JVM** — closures, dynamic scoping, non-local jumps, `eval`
 
-**300+ modules bundled inside the JAR:**
-```text
-%INC: 'Data/Dumper.pm' =>
-  'file:/path/to/perlonjava.jar!/lib/Data/Dumper.pm'
-```
-
----
-
-## XSLoader: Java Instead of C
-
-- Loads **Java extensions** instead of C shared libraries
-- **jnr-posix** replaces XS for native POSIX calls
-- No C compiler needed
-
-Note:
-Java equivalents are easier to write and maintain than C/XS. The same API surface is exposed to Perl code.
-
----
-
-# Section 4: Integration & Future
-
----
-
-## JSR-223: Embed Perl in Java
-
-JSR-223 is the standard Java scripting API (JDK since Java 6).
-
-```java
-ScriptEngineManager manager = new ScriptEngineManager();
-ScriptEngine perl = manager.getEngineByName("perl");
-
-perl.put("data", myJavaObject);
-Object result = perl.eval("process_data($data)");
-```
-
-**Bidirectional:** Java ↔ Perl seamlessly.
-
-**Use case:** Embed legacy Perl scripts in a modern Java application without rewriting them.
+Perl was never designed for the JVM — but careful engineering makes it work.
 
 ---
 
-## Future Targets
-
-**Current:** Standard JVM (HotSpot)
-
-1. **GraalVM** — native executables, instant startup
-2. **Android DEX** — Perl on mobile devices
-
-The Internal VM is key — custom bytecode is portable to any JVM derivative.
-
-Note:
-Dual backend matters beyond performance. GraalVM gives standalone executables. Android DEX converts JVM to Dalvik bytecode.
-
----
-
-## Interactive Debugger
-
-**Invoke with `-d` flag:** `./jperl -d script.pl`
-
-| Command | Action |
-|---------|--------|
-| `n` | Step over (next line) |
-| `s` | Step into subroutine |
-| `r` | Step out (return) |
-| `c` | Continue to breakpoint |
-| `b 42` | Set breakpoint at line 42 |
-| `l` | List source around current line |
-| `T` | Stack trace |
-| `p $var` | Print variable value |
-
-Supports `$DB::single`, `%DB::sub`, custom `DB::DB` — compatible with Perl's debugger API.
-
-Note:
-Debugger uses Internal VM (forced with -d). DEBUG opcodes inserted at each statement. DebugHooks handles breakpoints, command parsing, and eval in current scope. PERL5DB supported for custom debuggers.
-
----
-
-## Current Limitations
-
-**JVM-incompatible:**
-- `fork` — not available on JVM
-- `DESTROY` — JVM uses non-deterministic GC
-- Threading — not yet implemented
-
-**Partially implemented:**
-- Some regex features, taint checks
-
-Note:
-Workarounds: jnr-posix for native access, Java threading APIs, file auto-close at exit. XS modules use Java equivalents.
-
----
-
-## Roadmap
-
-**Stable now:** JVM backend, Perl class features, IPC, sockets, interactive debugger
-
-**In progress:** Internal VM optimization, eval STRING performance
-
-**Next:** More compatible regex engine, additional debugger features
-
----
-
-# Closing
-
----
-
-## Perl Was Never Designed to Run on the JVM
-
-We made it work anyway — and made it **fast**.
-
-<span class="metric">~200,000 tests</span> · <span class="metric">392 files</span> · <span class="metric">5,741 commits</span>
-
-No formal spec exists. The tests **are** the specification.
-
-Note:
-This is test-driven development at its most extreme — tests define the language behavior.
-
----
-
-## Get Involved
-
-**GitHub:** github.com/fglock/PerlOnJava · **License:** Artistic 2.0
-
-- **Test** your scripts and report issues
-- **Port** CPAN modules
-- **Contribute** to core development
-
----
-
-## Thank You!
-
-**Special thanks to:**
+## What's Next: Part 3
 
-- **Larry Wall** — for creating Perl
-- **Perl test writers** — tests that define Perl's behavior
-- **Perl community** — for decades of innovation
-- **Prior pioneers** — JPL, perljvm, Perlito5
+**Integration, tooling, and roadmap:**
 
-**Questions?** → github.com/fglock/PerlOnJava/issues
+- Module loading and XSLoader
+- JSR-223: embedding Perl in Java
+- Interactive debugger
+- Future targets and roadmap
 
 ---