* COMMENT -*- mode: org -*-
#+Date: 2015-09-07
Time-stamp: <2015-09-07>

* 2015-09-07 write an apache2 access log parser in haskell.
practice using attoparsec.

10.21.176.7 - - [06/Sep/2015:06:26:02 +0800] "GET /v3/auth/tokens HTTP/1.1" 200 3997 "-" "-" 1011068
10.21.176.7 - - [06/Sep/2015:06:26:03 +0800] "GET /v3/auth/tokens HTTP/1.1" 200 4065 "-" "python-requests/2.2.1 CPython/2.7.3 Linux/3.13.0-38-generic" 5481
10.21.176.7 - - [06/Sep/2015:06:26:02 +0800] "GET /v2.0/tokens/e79ce79f16f04852a7d981009090fb75 HTTP/1.1" 200 2692 "-" "python-requests/2.2.1 CPython/2.7.3 Linux/3.13.0-40-generic" 879293
10.21.176.7 - - [06/Sep/2015:06:50:02 +0800] "POST /v2.0/tokens HTTP/1.1" 200 2615 "-" "python-neutronclient" 211874238

allow me to
- filter slow requests.
  -gt 2s

- find most frequent request.
  order by request verb and path frequency.
  print out first N most frequently viewed verb&path.

  get top 5 request
  cd /var/log/apache2
  ~sylecn/a2p -i access.keystone.admin.ssl.log -t 5

  stack build && stack exec a2p -- -i ~/d/t1 --top 1

  output for user:
  M%   COUNT     POST xxx
  M%   COUNT     GET xxx

  return value for this
  Map (verb, path) count
  Order by count.
  get first N.

  filter the map before converting to alist.

  - how to do insert or update? when doing counting.
    you have to use lookup. it's probably not thread safe.

  - works perfectly on production access log.

how to use the parser?

a2p -i ACCESS_LOG_FILE [-gt 20s]

- created project at ~/haskell/apache2-log-parser

- let's test parse efficiency first.
  print a summary line:
  parsed N input lines in M seconds. (P line/s)

  - support -f param and parsing lines from file.

  - can't convert duration to Int easily.

- Applicative is powerful. I didn't know about <* before.
  logParser :: Parser [AccessLog]
  logParser = many $ parseLine <* endOfLine

- how to use int type param in optparse-applicative?

  haskell - How to use Int with the optparse-applicative inside a data constructor? - Stack Overflow
  http://stackoverflow.com/questions/31636962/how-to-use-int-with-the-optparse-applicative-inside-a-data-constructor