Generating queries can be quadratic in time

[lpsmith]

Possibly related to #39, it appears as though query generation is quadratic when it involves escaping via libpq's escapeStringConn command. Here's a simple test case that reproduces the behavior:

{-# LANGUAGE OverloadedStrings, BangPatterns #-}

import Database.PostgreSQL.Simple
import System.Environment

main = do
   (nstr:_) <- getArgs
   let n = read nstr :: Int
   c <- connectPostgreSQL ""
   !q <- formatMany c "insert into foo values (?,?)" [ (i,show i) | i <- [1..n] ]
   close c

Note that this isn't a problem if you replace show i by say, i^2 or whatnot.

[lpsmith]

Ok, I did a bit of information gathering and discovered that mysql-simple does not exhibit this bug, while even version 0.0 of postgresql-simple does. I find this slightly surprising as I didn't change Bryan's query generation code too much; I was expecting this problem to have either been introduced sometime after the first release or inherited from mysql-simple, but clearly I did something wrong in the initial port.

[lpsmith]

Whatever the problem is, at least some of it appears to be inside either the libpq C library and/or the postgresql-libpq haskell binding. This appears to exhibit non-linear behavior, though perhaps not quite quadratic, as doubling the number of escaped strings takes about 3x longer:

{-# LANGUAGE OverloadedStrings #-}

import qualified Database.PostgreSQL.LibPQ as PQ
import System.Environment
import qualified Data.ByteString.Char8 as B

main = do
    (nstr : _) <- getArgs
    let n = read nstr :: Int
    c <- PQ.connectdb ""
    stat <- PQ.status c
    case stat of
      PQ.ConnectionOk -> do
          xs <- loop c n []
          print (xs == xs)
          PQ.finish c
      _ -> fail "connection failed"

loop c n acc =
  if n == 0
  then return acc
  else do
    a <- PQ.escapeStringConn c "test"
    loop c (n-1) (a:acc)

[lpsmith]

Ok, I fixed up postgresql-libpq's binding to escapeStringConn, and the performance looks a lot more reasonable. I'm going to call this bug closed for now.