TODO
author Kristian H?gsberg <krh@redhat.com>
Tue Sep 04 23:52:59 2007 -0400 (2007-09-04)
changeset 6 4eeed5fbe6b7
parent 1 38be5ee4d231
child 8 7820b7d94662
permissions -rw-r--r--
Factor out array code.
krh@1
     1
- pkg manifest is list of files
krh@1
     2
krh@1
     3
	/usr/bin/bash 1321321372198798
krh@1
     4
krh@1
     5
  plus provides, requires and version?
krh@1
     6
krh@1
     7
- keep history of installed packages/journal of package transaction,
krh@1
     8
  so we can roll back to yesterday, or see what got installed in the
krh@1
     9
  latest yum update.
krh@1
    10
krh@1
    11
- we build a cache of the currently installed set to service
krh@1
    12
  dependency inquiries fast:
krh@1
    13
krh@1
    14
	map from property to pkg (as hash) providing it
krh@1
    15
	map from property to pkgs requiring it
krh@1
    16
	map from pkg name to manifest
krh@1
    17
	map from string to string pool index
krh@1
    18
krh@1
    19
	no implicit provides? not even pkgname?
krh@1
    20
krh@1
    21
- properties are strings, stored in a string table
krh@1
    22
krh@1
    23
- on disk maps are binary files of (string table index, hash) pairs
krh@1
    24
krh@1
    25
- at run time, we mmap the map, and keep changes in memory in a splay
krh@1
    26
  tree or similar.  if searching the splay tree fails we punt to the
krh@1
    27
  mmap.  once the transaction is done, we merge the map and the splay
krh@1
    28
  tree and write it back out.
krh@1
    29
krh@1
    30
- the on-disk string pool is sorted and we keep a list of indices into
krh@1
    31
  the string pool in sorted order so we can bsearch the list with a
krh@1
    32
  string to get its string pool index.  maybe a hash table is better,
krh@1
    33
  less I/O as we will expect to find the string within the block we
krh@1
    34
  look up with the hash function.
krh@1
    35
krh@5
    36
- represent all files as a breadth first traversal of the tree of all
krh@5
    37
  files.  each entry has its name (string pool index), the number of
krh@5
    38
  immediate children, total number of children, and owning package.
krh@5
    39
  for files both these numbers are zero.  a file is identified by its
krh@5
    40
  index in this flattened tree.
krh@5
    41
krh@5
    42
  to get the file name from an index, we search through the list.  by
krh@5
    43
  summing up the number of children, we know when to skip a directory
krh@5
    44
  and when to descend into one.  as we go we accumulate the path
krh@5
    45
  elements.
krh@5
    46
krh@5
    47
  hmm, dropping number of immediate children and using a sentinel drops
krh@5
    48
  a word from every entry.
krh@5
    49
krh@1
    50
- signed pkgs
krh@1
    51
- gzip pkg xml files somehow?