[RFC] branch --bind

[Stephen J. Turnbull]

Danny van Heumen writes:
 > John Arbash Meinel wrote:
 > > [snip]
 > > So *IMO* an optimal setup would be to get rid of heavyweight checkout as
 > > a concept. And stick with

I don't see this as an good answer to A Budden, who has a use case for
the *concept*.  It seems to me that it should be possible to
*automatically* implement the "heavyweight checkout = bound branch +
separate lightweight checkout" strategy.  This could be done by
convention (eg, reserve the .bzr-bound-branches subdirectory of the
shared repo for bound branches, what a surprise!) and then do the
equivalent of

heavyweight-checkout () {

    # tell user what's already there
    if test "$1"="--show"; then
        echo $MY_SHARED_REPO/.bzr-bound-branches
        ls $MY_SHARED_REPO/.bzr-bound-branches
        return
    fi

    # make a new heavyweight checkout
    # FIXME: think about allowing reuse of bound branches of the same
    # source
    source_branch=$1
    if test -n "$2"; then
        checkout_location=$2
    else
        checkout_location=`pwd`/`basename $source_branch`
    fi
    if test -n "$3"; then
        shared_repo=$3
    else
        shared_repo=$MY_SHARED_REPO
    bound_branch=.bzr-bound-branches/`basename $checkout_location`
    pushd $shared_repo
    bzr branch $source_branch $bound_branch
    cd $bound_branch
    bzr bind $source_branch
    popd
    bzr checkout --lightweight $bound_branch $checkout_location
}

(probably as a plugin or even use the above as a contrib shell
script).

Then you can reduce the *implementation* "primitive concepts" to

 > > Repository
 > > Branch
 > > Checkout

while still offering the heavyweight checkout concept to users.  I
don't know if this implementation would be acceptable to A Budden,
though.

Question: in the above classification, is "checkout" identical to
"version-controlled working tree" from the user's standpoint?
(Obviously for implementation the working tree is "everything but the
.bzr subtree".)

 > +1
 > 
 > I have more than once done a checkout, only to hit my head against the 
 > wall during the time all the history is being transferred :P.
 > For some reason I always expect a lightweight checkout (otherwise I 
 > would've just asked for a branch), only to find out during the process 
 > that it isn't.

Sure, but that's easy to fix by changing the default for "bzr
checkout".  (Technically easy, the cost of user transitions is of
course potentially large.)