~bzr-pqm/bzr/bzr.dev

=============================

Bazaar Architectural Overview

=============================

This document describes the key classes and concepts within Bazaar.  It is

intended to be useful to people working on the Bazaar codebase, or to

people writing plugins.  People writing plugins may also like to read the 

guide to `Integrating with Bazaar <integrating.html>`_ for some specific

recipes.

If you have any questions, or if something seems to be incorrect, unclear

or missing, please talk to us in ``irc://irc.freenode.net/#bzr``, or write

to the Bazaar mailing list.  

Using bzrlib

############

Within bzr

==========

When using bzrlib within the ``bzr`` program (for instance as a bzr

plugin), bzrlib's global state is already available for use.

From outside bzr

================

To use bzrlib outside of ``bzr`` some global state needs to be setup.

bzrlib needs ways to handle user input, passwords, a place to emit

progress bars, logging setup appropriately for your program. The easiest

way to set all this up in the same fashion ``bzr`` does is to call

``bzrlib.initialize``. This returns a context manager within which bzrlib

functions will work correctly. See the pydoc for ``bzrlib.initialize`` for

more information. In Python 2.4 the ``with`` keyword is not supported and

so you need to use the context manager manually::

  # This sets up your ~/.bzr.log, ui factory and so on and so forth. It is

  # not safe to use as a doctest.

  library_state = bzrlib.initialize()

  library_state.__enter__()

  try:

      pass

      # do stuff here

  finally:

      library_state.__exit__(None, None, None)

Core classes

############

Transport

=========

The ``Transport`` layer handles access to local or remote directories.

Each Transport object acts as a logical connection to a particular

directory, and it allows various operations on files within it.  You can

*clone* a transport to get a new Transport connected to a subdirectory or

parent directory.

Transports are not used for access to the working tree.  At present

working trees are always local and they are accessed through the regular

Python file I/O mechanisms.

Filenames vs URLs

-----------------

Transports work in terms of URLs.  Take note that URLs are by definition

only ASCII - the decision of how to encode a Unicode string into a URL

must be taken at a higher level, typically in the Store.  (Note that

Stores also escape filenames which cannot be safely stored on all

filesystems, but this is a different level.)

The main reason for this is that it's not possible to safely roundtrip a

URL into Unicode and then back into the same URL.  The URL standard

gives a way to represent non-ASCII bytes in ASCII (as %-escapes), but

doesn't say how those bytes represent non-ASCII characters.  (They're not

guaranteed to be UTF-8 -- that is common but doesn't happen everywhere.)

For example, if the user enters the URL ``http://example/%e0``, there's no

way to tell whether that character represents "latin small letter a with

grave" in iso-8859-1, or "latin small letter r with acute" in iso-8859-2,

or malformed UTF-8.  So we can't convert the URL to Unicode reliably.

Equally problematic is if we're given a URL-like string containing

(unescaped) non-ASCII characters (such as the accented a).  We can't be

sure how to convert that to a valid (i.e. ASCII-only) URL, because we

don't know what encoding the server expects for those characters.

(Although it is not totally reliable, we might still accept these and

assume that they should be put into UTF-8.)

A similar edge case is that the URL ``http://foo/sweet%2Fsour`` contains

one directory component whose name is "sweet/sour".  The escaped slash is

not a directory separator, but if we try to convert the URL to a regular

Unicode path, this information will be lost.

This implies that Transports must natively deal with URLs.  For simplicity

they *only* deal with URLs; conversion of other strings to URLs is done

elsewhere.  Information that Transports return, such as from ``list_dir``,

is also in the form of URL components.

WorkingTree

===========

A workingtree is a special type of Tree that's associated with a working

directory on disk, where the user can directly modify the files. 

Responsibilities:

 * Maintaining a WorkingTree on disk at a file path.

 * Maintaining the basis inventory (the inventory of the last commit done)

 * Maintaining the working inventory.

 * Maintaining the pending merges list.

 * Maintaining the stat cache.

 * Maintaining the last revision the working tree was updated to.

 * Knows where its Branch is located.

Dependencies:

 * a Branch

 * an MutableInventory

 * local access to the working tree

Branch

======

A Branch is a key user concept - its a single line of history that one or

more people have been committing to. 

A Branch is responsible for:

 * Holding user preferences that are set in a Branch.

 * Holding the 'tip': the last revision to be committed to this Branch. (And the revno of that revision.)

 * Knowing how to open the Repository that holds its history.

 * Allowing write locks to be taken out to prevent concurrent alterations to the branch.

Depends on:

 * URL access to its base directory.

 * A Transport to access its files.

 * A Repository to hold its history.

Repository

==========

Repositories store committed history: file texts, revisions, inventories,

and graph relationships between them.  A repository holds a bag of

revision data that can be pointed to by various branches:

 * Maintains storage of various history data at a URL:

   * Revisions (Must have a matching inventory)

   * Digital Signatures

   * Inventories for each Revision. (Must have all the file texts available).

   * File texts

 * Synchronizes concurrent access to the repository by different

   processes.  (Most repository implementations use a physical 

   mutex only for a short period, and effectively support multiple readers

   and writers.)

Stacked Repositories

--------------------

A repository can be configured to refer to a list of "fallback"

repositories.  If a particular revision is not present in the original

repository, it refers the query to the fallbacks.

Compression deltas don't span physical repository boundaries.  So the

first commit to a new, empty repository with fallback repositories will

store a full text of the inventory, and of every new file text.

At runtime, repository stacking is actually configured by the branch, not

the repository.  So doing ``a_bzrdir.open_repository()``

gets you just the single physical repository, while

``a_bzrdir.open_branch().repository`` gets one configured with a stacking.

Therefore, to permanently change the fallback repository stored on disk,

you must use ``Branch.set_stacked_on_url``.

Changing away from an existing stacked-on URL will copy across any

necessary history so that the repository remains usable.

A repository opened from an HPSS server is never stacked on the server

side, because this could cause complexity or security problems with the

server acting as a proxy for the client.  Instead, the branch on the

server exposes the stacked-on URL and the client can open that.

..

   vim: ft=rst tw=74 ai