Hope Lexical analyzer is working

I wrote quite a significant part of lexical analyzer for Hope (http://kenai.com/projects/hope) (Haskell Oriented Python Environment). Its primarily based on Parsec library. Currently the code is organized as follows. Interpreter/src directory contains a folder PyLex. PyLex package contains various files which implement the lexical analysis for Hope. The main module is PyLex.Lexer. You can start with the function pyLexer which a Parsec based parser and returns a list of tokens. Each token is a pair of (SourcePos, PyToken). PyToken data-type represents various types of tokens supported in Hope's lexical analyzer. The tokens have been defined in PyLex.Tokens file. The implementation of parsing for these tokens is spread over multiple modules (Identifier, Number, Operator, String).

I faced quite a few issues in reaching up to this stage. I read about Parsec from a number of sources.

• Parsec tutorial and reference guide
• Real world Haskell (online book)
• Develop your own scheme interpreter in 24 hours

It took me quite a while in getting a good feel about how to use Parsec. I would try to write about some of the things I learnt. The import statements in the Parsec tutorial are outdated. Correct imports are:

import Text.ParserCombinators.Parsec

import qualified Text.ParserCombinators.Parsec.Token as PT

import qualified Text.ParserCombinators.Parsec.Language as PL

import qualified Text.ParserCombinators.Parsec.Pos as PP

(I used qualified imports, you may not need them).

The tutorial says that 2 stage lexical analysis and parsing is the right way to go. It says that the lexical analyzer should return a list of tokens and their source positions. It took me quite a while to figure out how to extract the source position inside the character parser when required. And it turns out to be quite easy. Here is an example (sorry for improper indentation):

-- unsupported character inside a python file

pyUnsupported :: Parser Token

pyUnsupported = do pos <- getPosition

c <- anyChar

return (pos, (PyUnsupported c))

"unsupported character"

Important thing to notice is the pos <- getPosition call. This is the way we can extract the source position.

A nice function I wrote to convert a list of digits in base N to an integer. Here it is:

-- converts a string of digits to an integer for a given base

-- n is an accumulator which should be initialized with 0

-- example : convToDecimal [1,1,0,0] 2 0

convToDecimal :: [Int]->Integer->Integer->Integer

convToDecimal [] base n = n

convToDecimal (x:xs) base n = convToDecimal xs base (n * base + fromIntegral x)

The syntax for strings in Python is really quite sophisticated. You have got short strings and long strings ("""abc""" types). And both strings can have single quotes or double quotes. And then you have raw strings. And finally unicode strings and byte strings have different prefixes. Off course a lot of escape characters have to be especially handled inside the strings. My current implementation is not comprehensive but it covers pretty much all normal unicode strings. It doesn't yet cover raw strings and byte strings. I would cover them later. First I got to get the interpreter working. I don't really know if my parser is really great. I don't have enough Haskell experience yet to be able to say that. Still I am happy I got something working finally.

Currently going through Modern Compiler Design book. I hope to get the parser working and ASTs generated. May be then I will be able to translate them into Python Byte Codes and then get an interpreter virtual machine working! Its a lot of work and I am not a computer scientist :)

PyCon India 2009 My impressions

I attended 1st day (26 September 2009) of the PyCon India 2009 conference. It was first Python Conference in India and I felt inclined enough to go all over to Bangalore from my home place Noida and attend the conference.

The conference was organized at IISc Bangalore. From the main gate to reach the conference area was easy as proper directions had been placed all over. Registration was a bit chaotic. There were not enough desks to handle the huge crowd waiting to be registered. I guess it was still quite satisfactory process considering the fact that it was first such event in India and there were not enough funds (I guess) to manage the event professionally. Since I had pre-registered, I had to pay just 200 rupees for attendance. They provided us with schedule, a T shirt, a notebook, a pen and off course a badge which was quite good as a conference kit. And the lunch and tea breaks were excellent.

There were about 600+ delegates participating in the conference. The profile varied between students, professors, independent consultants, professional programmers etc.

The first session by Prabhu Ramchandran (My adventures) was quite nice. It was interesting to see how an open source project Mayavi had been run successfully over the years out of India. Since there were three parallel tracks going on, hence I had to identify the sessions I wished to attend.

Some of the sessions had a lot of issues related to presentation. The presentation on zc.buildout was quite boring. The presenter spent too much time just discussing what is isolation and repeatability that there wasn't much time left for actual demonstration. Similarly, the session on DJango turned out to be quite boring. There were three presenters taking turns in the talk. They spent a lot of time preaching about DJango, like what it is, why it is great, etc. etc. They also spend a lot of time typing the code which they wished to demonstrate. I guess one should be very careful in using the 45 minutes available during a conference time and make the best use of it as possible. One should avoid time wasters like typing code and fixing bugs during the talk itself. Another boring session was on Semantic web. The presenter was explaining a whole lot of difficult concepts. It seemed like a typical IIT lecture in which students preferred to sleep through or if possible bunk them.

Another interesting talk was 'Jython at Strand LS' by Janaki where the speaker explained how they used Jython at their company. Examples were short and sweet. Illustrations were easy to understand and the speaker had a great sense of humour. I was impressed to see the extensive application of Jython in a production quality platform like AVADIS (TM) at Strand.


Finally, my own talk on using ZPT and RML for generating PDF in Python applications came. I didn't expect too many to attend the talk but I was happy to see about 30+ ppl still continuing to sit for my session. Some of them left during the session which meant that they didn't really enjoy the content much. Others continued through the session and I guess they liked what I talked. I would leave it for others to comment on how good the session was.

There were tracks for lightning talks but I guess very few speakers were actually prepared to conduct them. There were lots of talks registered but ultimately many of them were not delivered.

Some of the things I feel that should be improved for future conferences:

• Expected audience of a talk should be more clearly explained
• Focus should be more on specific coding workshops rather than general talks
• Longer workshops on specific topics should be conducted if there are speakers available
• There was no peer review of presentation material. This should be extensively fixed.
• The work on presentation materials should actually finish 2-3 months before the conference at least so that peer reviews and revisions can be completed say about 2 weeks before actual conference.
• Scheduling of talks should happen through a more transparent process
• Registered delegates should receive notifications about talks (when its announced, when presentation materials are uploaded etc.) over mail as the process keeps on going before the actual conference
• The talks without presentation material should be rejected in the final schedule.
• Speakers should avoid wasting too much time talking about abstract concepts and spend more on illustrations. They should also avoid typing code during the session.
• Multi-speaker talks should be minimized

Hope Project at Kenai

Today I started an attempt to write an implementation of Python in Haskell. I have no idea whether I would really succeed in this. But I really want to give it a good try.
The site is hosted at http://kenai.com/projects/hope. I will continue to blog about this from this blog.

HOPE stands for Haskell Oriented Python Environment. At the moment, its just a dream for me and a very very big dream at that. But I hope that I will be able to bring out HOPE sometime !

Quicksort in Haskell

Here is my attempt at quicksort in Haskell

quicksort :: (Ord a) => [a] -> [a]
quicksort [] = []
quicksort (x:xs) =  smaller ++ [x] ++ bigger
                   where smaller = quicksort [y| y <- xs , y < x]
                         bigger = quicksort [y | y <- xs, y >= x]
      

So how does this work?

• [a] ->[a] means that quicksort takes a list as input and returns a list as output


• (Ord a) means that a belongs to a type class Ord (i.e. a is an ordered type)


• The first expression means that quicksort returns an empty list for an input empty list


• The second expression works as follows. The input list is broken into its head and tail (x, xs). We use x as the pivot element. all elements smaller than x in xs are identified and sorted using quicksort itself (recursively). All elements greater than or equal to x are sorted by quicksort itself (recursively). The final result is smaller + [x] + bigger.

So here are some examples:

*Main> quicksort [2, 1, 3]
[1,2,3]
*Main> quicksort "i don't believe you"
"   'bdeeeiilnootuvy"
*Main> quicksort "i love you"
"  eiloouvy"
*Main> quicksort [50, 40, 30 , 10, 20, 30, 40, 50, 20]
[10,20,20,30,30,40,40,50,50]
*Main>

testing code prettifier

Just learnt about how to prettify code inside Blogger.
So trying it out now.

print "Hello World"
for i in range(10):
    print i,
print

Well this works nicely.

For more information about how to set it up, please follow it at:

http://lukabloga.blogspot.com/2008/10/to-test-new-highlighting.html