PHP vs. Python Revisited

After some thought, I’ve decided that I won’t switch back to PHP. I’ve realized that the major roadblocks I’ve been running into recently have all been related to the way plugins hook into the braindump core. In fact, I’ve been spinning my wheels for months with issues related to braindump’s extensibility. Switching back to PHP will solve the two problems listed in the previous post (because of how PHP includes source code into one big namespace and sends stdout to the web) but I’ll inevitably run into more problems with extensibility in the future — trying to think ahead and allow for all the possible things a plugin might want to implement is just a huge time suck.

So Python isn’t my problem — my problem is language-independent and therefore wouldn’t be fixed by switching back to PHP (phew, didn’t want to have to do that). My problem — what’s prevented me from implementing any major new functionality in months — is that I’m spinning my wheels in thoughts about extensibility instead of actually implementing new functionality. Now extensibility and modularity are great, but doing them well just takes so much time and thought, as well as a well-defined idea of what the core should be: its boundaries, its implementation, etc. I don’t have this good of an idea yet, and I won’t for a long time. As the project page says, the purpose of braindump is to experiment; it’s not meant to be a polished product for mainstream use. Furthermore, even if I did build braindump to be polished and extensible, no one would write plugins. Let’s face it, I only know of one person ever even downloading and installing it.

So, I might as well stop spinning my wheels in extensibility issues and get moving! Some things I want to do:

• view and edit data from multiple pages in a table
• import metadata about iTunes songs from iTunes XML export files (already implemented in the PHP version)
• more comparison operators — right now there’s just ‘is’. I want ‘>’ and ‘<’ for numbers, maybe regular expression matching, maybe ‘before’ and ‘after’ for dates and even ‘north of’ ‘south of’ (etc) and ‘within x miles of’ for locations
• visualizations
  • Google Maps
  • SIMILE timeline
  • Graphs/Charts (probably with the Google Chart API
• interface enhancements (autocompletion, etc) — making the interface as powerful and intuitive as possible

…and I’m sure I’ll think of more. I’m excited about this new mindset. Hopefully it’ll help me make cool stuff faster.

PHP vs. Python

I’m running into a lot of problems with my python port of braindump — things that were easy with PHP are slowing me down a lot. I’m considering switching back. Here’s my brainstorm: (more for me than anyone else)

problems I’m having with python:

• importing
  • importing plugins was solved, but importing modules from mako is another nightmare because it imports them from mako’s runtime namespace.
  • finding the file and managing the namespace is so easy in PHP!!

• plugins
  • instead of just printing to stdout, they have to do a context.write(). god knows how and when I’ll get the context object

• hosting
  • won’t run on GAE because graphstore reads & writes with pure SQL!
  • PHP hosting is so damn easy!!

• HTTP stuff
  • cookies, sessions, GET/POST vars, headers all built into PHP; need some python framework
  • static files are a breeze because they’re right there next to the PHP — don’t need to pass through the program or anything and can be gzipped and stuff with the server

reasons why I switched:

• libraries!
  • diff match patch
    • diff not needed cuz don’t really do revisioning; maybe fuzzy match for search/autocompletion…
  • montylingua, wordnet, conceptnet
    • remains to be seen how these would be integrated. besides, python scripts could always download, process, and upload with XML-RPC…
  • PHP pretty much has everything I need anyway
    • templating (Twig), YAML, JSON, XML-RPC, XMPP, strtotime, textile, syntax highlighting, image manipulation, etc

• language itself
  • python syntax is undeniably better. I’ll miss list comprehensions, no semicolons, and whitespace instead of curly braces… but PHP is good enough.
  • the PHP console never really worked for me…

• standaloneness
  • ideally, you would download braindump, unzip it, start it up with python braindump.py, and go to http://localhost:8080/. In reality, you would have to easy_install mako and cherrypy and stuff, but that’s still easier than installing a PHP-enabled web server

• ground-up design of web architecture
  • WSGI, FastCGI etc all designed, not built up ad hoc like PHP
  • PHP was meant to just be embedded; MVC and routing and stuff are just done with (but the hacks work)

what I’ll do if I switch back to PHP

• integrate Twig. I don’t think I can go back to living without a templating engine. they’re so nice.
• de-kludgify my framework’s whole routing system
• restructure the directories to make more sense
• work on graph.php
  • make it into a class that is instantiated (new Graph('path/to/database.sqlite');)
  • implement filtering with multiple conditions
  • use pure SQL instead of going through my kludgy object-oriented database library? exception/null handling less complicated…

The Blank Slate

For the past year or so, off-and-on, I’ve been reading The Blank Slate by Stephen Pinker. It’s really, really good. Pinker’s central thesis is that incorrect theories of human nature are at the root of many of society’s most pressing problems, and that we can move forward by basing our policies and behaviors on a new theory of human nature derived from scientific investigation, rather than armchair philosophy. To flesh out this argument, Pinker must move through some of the most controversial territory there is, but he consistently makes the journey with refreshing rationality and intellectual honesty, and a complete lack of mud-slinging.

Although he steers clear of the God vs. Science debate, Pinker makes it clear that the Judeo-Christian theory of human nature (dominant in this country and much of the world) is outdated and incorrect, and that we need to move past it. That’s old news to most intellectuals, but Pinker also pans the Enlightenment view, commonly held among intellectuals, that at birth the human mind is a “blank slate” to be written upon. (the book’s title is not an endorsement of this view, but an example of “The Modern Denial of Human Nature”, which is the book’s subtitle) Pinker then charts a course through the Nature vs. Nurture debate using the findings of evolutionary psychology, behavioral genetics, neuroscience, and cognitive science (“bridges between biology and culture” and “the new sciences of human nature”, he calls them) to place the line somewhere between the two extremes: all people have the same innate mechanisms, which allow them to learn an infinite number of different things and behave in an infinite number of different ways.

He then shows how this new theory of human nature, grounded in science (and therefore not beyond question and disprovement, as are the religious theories) can help us make progress on the most difficult issues of our society, such as politics, violence, abortion, stem cells, and education.

I’m not done yet — the book is so packed with big ideas and interesting research that it’ll be a long time until I’ve fully digested it. But I won’t miss the rest, and I’ll probably turn to it when the presidential campaign and war and American society become too much — the book’s rationality, intellectual honesty, and complete lack of mud-slinging on the issues that matter is just what the doctor ordered — for me, for this country and for this world.