package VCP::Source ; =head1 NAME VCP::Source - A base class for repository sources =head1 SYNOPSIS =head1 DESCRIPTION =head1 OPTIONS =over =item --bootstrap --bootstrap=pattern Forces all files matching the given shell regular expression (may use wildcards like "*", "?", and "...") to have their first revisions transferred as complete copies instead of deltas. This is useful when you want to transfer a revision other than the first revision as the first revision in the target repository. It is also useful when you want to skip some revisions in the target repository (although the L<Map filter|VCP::Filter::map> has superceded this use). =item --continue Tells VCP to continue where it left off from last time. This will not detect new branches of already transferred revisions (this limitation should be lifted, but results in an expensive rescan of metadata), but will detect updates to already transferred revisions. =back =cut $VERSION = 0.1 ; use strict ; use UNIVERSAL qw( isa ) ; use VCP::Debug qw( :debug ) ; use VCP::Logger qw( lg BUG ); use base 'VCP::Driver' ; use fields ( 'BOOTSTRAP', ## The raw option so we can regurgitate it 'BOOTSTRAP_REGEXPS', ## Determines what files are in bootstrap mode. 'DEST', 'CONTINUE', ## Set if we're resuming from the prior ## copy operation, if there is one. This causes ## us to determine a minimum rev by asking the ## destination what it's seen on a given filebranch 'SENT_REV_COUNT', ## Number of revs sent ## Turns out that most real repositories (ie not RevML, at least) ## are most easily scanned in reverse chronological order. Keeping ## the last revision or the last revision by filebranch is handy in ## these cases. 'LAST_REV', ## The rev that was last sent 'LAST_REV_BY_FILEBRANCH', ## The last sent 'SEEN_IDS', ## IDs of those revisions set already. 'SEND_REV_WHEN_PREVIOUS_ID_SET', ) ; sub init { my VCP::Source $self = shift; $self->bootstrap( $self->{BOOTSTRAP} ); $self->{SENT_REV_COUNT} = 0; $self->{SEND_REV_WHEN_PREVIOUS_ID_SET} = {}; } ############################################################################### =head1 SUBCLASSING This class uses the fields pragma, so you'll need to use base and possibly fields in any subclasses. See L<VCP::Plugin> for methods often needed in subclasses. =head2 Subclass utility API =over =item options_spec Adds common VCP::Source options to whatever options VCP::Plugin parses: =cut sub options_spec { my VCP::Source $self = shift; return ( $self->SUPER::options_spec, "bootstrap|b=s" => \$self->{BOOTSTRAP}, "continue" => \$self->{CONTINUE}, "rev-root=s" => \$self->{REV_ROOT}, ); } =item dest Sets/Gets a reference to the VCP::Dest object. The source uses this to call handle_header(), handle_rev(), and handle_end() methods. =cut sub dest { my VCP::Source $self = shift ; $self->{DEST} = shift if @_ ; return $self->{DEST} ; } =item continue Sets/Gets the CONTINUE field (which the user sets via the --continue flag) =cut sub continue { my VCP::Source $self = shift ; $self->{CONTINUE} = shift if @_ ; return $self->{CONTINUE} ; } =item real_source Returns the reference to be used when sending revisions to the destination. Each revision has a pointer to the source that sends it so that filters and destinations can call get_source_file(). Most sources return $self; Sources that spool data, such as VCP::Source::metadb, need to specify a real source. They do so by overloading this method. VCP::Source::revml does not do this, as it supplies a get_source_file(). =cut sub real_source { return shift; } =item send_rev $self->send_rev( $r ); As the revisions are scanned, the source sends them downstream to the dest using this method. Sources should not retain references to revisions, they should copy them if needed, or better yet, copy *just* the required metadata as it is needed. This is a requirement so that filters may alter the revisions without affecting the source's logic. This updates last_rev and last_rev_for_filebranch. =cut sub _send_rev { my VCP::Source $self = shift ; my ( $r ) = @_; debug "_send_rev: ", $r->id if debugging; $r->set_source( $self->real_source ); ++$self->{SENT_REV_COUNT}; delete $self->{SEND_REV_WHEN_PREVIOUS_ID_SET}->{int $r}; $self->dest->handle_rev( $r ) if $self->dest; } sub send_rev { my VCP::Source $self = shift ; my ( $r ) = @_; $self->{LAST_REV} = $r; $self->{LAST_REV_BY_FILEBRANCH}->{$r->source_filebranch_id} = $r; $self->{SEEN_IDS}->{$r->id} = undef; $self->_send_rev( $r ); } =item queue_rev Some revs can't be sent immediately. They get queued. This updates last_rev and last_rev_for_filebranch. =cut sub queue_rev { my VCP::Source $self = shift ; my ( $r ) = @_; debug "queue_rev: ", $r->id if debugging; $self->{LAST_REV} = $r; $self->{LAST_REV_BY_FILEBRANCH}->{$r->source_filebranch_id} = $r; $self->{SEEN_IDS}->{$r->id} = undef; $self->revs->add( $r ); } =item queue_rev_until_previous_id_set Some revs can't be sent immediately. They get queued until their previous_id is set by set_last_rev_in_filebranch_previous_id(). This updates last_rev and last_rev_for_filebranch. =cut sub queue_rev_until_previous_id_set { my VCP::Source $self = shift ; my ( $r ) = @_; debug "queue_rev_until_previous_id_set: ", $r->id if debugging; $self->{LAST_REV} = $r; $self->{LAST_REV_BY_FILEBRANCH}->{$r->source_filebranch_id} = $r; $self->{SEEN_IDS}->{$r->id} = undef; $self->{SEND_REV_WHEN_PREVIOUS_ID_SET}->{int $r} = $r; } =item queued_rev $self->queued_rev( $id ); Returns a queued rev by id. Sources where revs can arrive willy-nilly, like VCP::Source::revml, queue up all revs and need to randomly access them. =cut sub queued_rev { my VCP::Source $self = shift ; return $self->revs->get( @_ ); } =item last_rev Returns the last revision sent or queued. =cut sub last_rev { my VCP::Source $self = shift ; my ( $r ) = @_; return $self->{LAST_REV}; } =item queued_revs Returns a list of all queued revs. Does not remove them from the queue. =cut sub queued_revs { my VCP::Source $self = shift; return $self->revs->get; } =item last_rev_for_filebranch $self->last_rev_for_filebranch( $filebranch_id ); Returns the last revision sent or queued on the indicated filebranch. =cut sub last_rev_for_filebranch { my VCP::Source $self = shift ; my ( $filebranch_id ) = @_; return $self->{LAST_REV_BY_FILEBRANCH}->{$filebranch_id}; } =item set_last_rev_in_filebranch_previous_id $self->set_last_rev_in_filebranch_previous_id( $r ); If there is a last_rev_for_filebranch for $r->filebranch_id, sets its previous_id to point to $r. This is useful for sources which scan in most-recent-first order. =cut sub set_last_rev_in_filebranch_previous_id { my VCP::Source $self = shift ; my ( $r ) = @_; BUG "\$r undef" unless defined $r; BUG "\$r = '$r'" unless ref $r; my $child_rev = $self->last_rev_for_filebranch( $r->source_filebranch_id ); if ( $child_rev ) { debug "setting ", $child_rev->id, "->previous_id to ", $r->id if debugging; $child_rev->previous_id( $r->id ); $self->_send_rev( $child_rev ) if $self->{SEND_REV_WHEN_PREVIOUS_ID_SET}->{int $child_rev}; } } =item send_rev_because_previous_id_was_set =cut sub send_rev_because_previous_id_was_set { my $self = shift; my ( $r ) = @_; $self->_send_rev( $r ); } =item last_revs_for_all_filebranches $self->last_revs_for_all_filebranches; Returns the last revision sent or queued on every filebranch =cut sub last_revs_for_all_filebranches { my VCP::Source $self = shift ; return values %{$self->{LAST_REV_BY_FILEBRANCH}}; } =item id_seen $self->id_seen( $id ); Returns true if the indicated id was sent or queued. =cut sub id_seen { my VCP::Source $self = shift; my ( $id ) = @_; return exists $self->{SEEN_IDS}->{$id}; } =item sent_rev_count Returns (does not set) the number of revs sent so far. =cut sub sent_rev_count { my VCP::Source $self = shift; return $self->{SENT_REV_COUNT}; } =item send_revs $self->send_revs; Removes and sends all revs accumulated so far. =cut sub send_revs { my VCP::Source $self = shift ; my ( $revs ) = @_; debug "sending revs" if debugging; $revs ||= $self->revs->remove_all; ## Oddly, we can't show the progress bar here because filters in the ## chain may accumulate revisions and sort them, so this is not a good ## metric. for my $i ( 0..$#$revs, ) { $self->send_rev( $revs->[$i] ); $revs->[$i] = undef; } for my $r ( values %{$self->{SEND_REV_WHEN_PREVIOUS_ID_SET}} ) { $self->send_rev( $r ); } $self->{SEND_REV_WHEN_PREVIOUS_ID_SET} = {}; } =back =head1 SUBCLASS OVERLOADS These methods should be overridded in any subclasses. =over =cut sub copy_revs { ## TODO: delete this (DEPRECATED) my VCP::Source $self = shift ; my ( $revs ) = @_; $self->send_revs; } =item get_source_file All sources must provide a way for the destination to fetch a revision. =cut sub get_source_file { my VCP::Source $self = shift; die $self, " does not overload get_source_file()\n"; } =item handle_header REQUIRED OVERLOAD. Subclasses must add all repository-specific info to the $header, at least including rep_type and rep_desc. $header->{rep_type} => 'p4', $self->p4( ['info'], \$header->{rep_desc} ) ; The subclass must pass the $header on to the dest: $self->dest->handle_header( $header ) if $self->dest; This may be called when dest is null to allow the source to initialize itself when it won't be scanning the real source. So the if $self->dest is important. That's not the case for copy_revs(). =cut sub handle_header { my VCP::Source $self = shift ; # my ( $header ) = @_ ; BUG "ERROR: handle_header not overloaded by class '", ref $self, "'. Oops.\n"; # if $self->can( 'handle_header' ) eq \&handle_header ; # $self->dest->handle_header( $header ) ; } =item handle_footer Not a required overload, as the footer carries no useful information at this time. Overriding methods must call this method to pass the $footer on: $self->SUPER::handle_footer( $footer ) ; =cut sub handle_footer { my VCP::Source $self = shift ; my ( $footer ) = @_ ; ## Release memory. $self->{LAST_REV_BY_FILEBRANCH} = undef; $self->{LAST_REV} = undef; $self->{SEEN_IDS} = undef; $self->{SEND_REV_WHEN_PREVIOUS_ID_SET} = undef; $self->revs->set; $self->dest->handle_footer( $footer ) ; } =item parse_time $time = $self->parse_time( $timestr ) ; Parses "[cc]YY/MM/DD[ HH[:MM[:SS]]]". Will add ability to use format strings in future. HH, MM, and SS are assumed to be 0 if not present. Returns a time suitable for feeding to localtime or gmtime. Assumes local system time, so no good for parsing times in revml, but that's not a common thing to need to do, so it's in VCP::Source::revml.pm. =cut { ## This routine is slow and gets called a *lot* with duplicate ## inputs, at least by VCP::Source::cvs, so we memoize it. my %cache; sub parse_time { my VCP::Source $self = shift ; my ( $timestr ) = @_ ; return $cache{$timestr} ||= do { ## TODO: Get parser context here & give file, line, and column. ## filename and rev too, while we're scheduling more work for ## the future. BUG "Malformed time value $timestr\n" unless $timestr =~ /^(\d\d)?\d?\d(\D\d?\d){2,5}/ ; my @f = split( /\D/, $timestr ) ; if ( length $f[0] <= 2 && $f[0] <= 12 && ( length $f[2] == 4 || $f[2] > 12 || "0" eq substr( $f[2], 0, 1 ) ) ) { ## Must be MM/DD/YY, or MM/DD/YYYY. timelocal() needs ## YY(YY)?/MM/DD splice @f, 0, 3, ( $f[2], $f[0], $f[1] ); } --$f[1] ; # Month of year needs to be 0..11 push @f, ( 0 ) x ( 6 - @f ) ; require Time::Local; my $t = eval { Time::Local::timelocal( reverse @f ) }; BUG $@ unless defined $t; return $t; } } } =item bootstrap Sets (and parses) or gets the bootstrap spec. Can be called plain: $self->bootstrap( $bootstrap_spec ) ; See the command line documentation for the format of $bootstrap_spec. =cut sub bootstrap { my VCP::Source $self = shift ; if ( @_ ) { my ( $val ) = @_ ; $self->{BOOTSTRAP} = $val; $self->{BOOTSTRAP_REGEXPS} = [ defined $val ? map $self->compile_path_re( $_ ), split /,+/, $val : () ]; } return $self->{BOOTSTRAP}; } =item is_bootstrap_mode ... if $self->is_bootstrap_mode( $file ) ; Compares the filename passed in against the list of bootstrap regular expressions set by L</bootstrap>. The file should be in a format similar to the command line spec for whatever repository is passed in, and not relative to rev_root, so "//depot/foo/bar" for p4, or "module/foo/bar" for cvs. This is typically called in the subbase class only after looking at the revision number to see if it is a first revision (in which case the subclass should automatically put it in bootstrap mode). =cut sub is_bootstrap_mode { my VCP::Source $self = shift ; my ( $file ) = @_ ; my $result = grep $file =~ $_, @{$self->{BOOTSTRAP_REGEXPS}} ; lg( "$file ", ( $result ? "=~ " : "!~ " ), "[ ", join( ', ', map "qr/$_/", @{$self->{BOOTSTRAP_REGEXPS}} ), " ] (", ( $result ? "not in " : "in " ), "bootstrap mode)" ) if debugging; return $result ; } =back =head1 COPYRIGHT Copyright 2000, Perforce Software, Inc. All Rights Reserved. This module and the VCP package are licensed according to the terms given in the file LICENSE accompanying this distribution, a copy of which is included in L<vcp>. =head1 AUTHOR Barrie Slaymaker <barries@slaysys.com> =cut 1
# | Change | User | Description | Committed | |
---|---|---|---|---|---|
#46 | 5404 | Barrie Slaymaker |
- SVN support added - Makefile gives clearer notices about missing optional prereqs. - VCP::Filter::labelmap and VCP::Filter::map: <<skip>> replaces deprecated <<delete>> to be clearer that no revisions are deleted from either repository but some just are skipped and not inserted. - VCP::Filter::map: support added for SVN-like branch labels - VCP::Source: support added for ISO8601 timestamps emitted by SVN. |
||
#45 | 5082 | Barrie Slaymaker |
- VCP::Source tells VCP::Rev to uncache the source to allow the source instance to be DESTROYed and thus clean up its working files. |
||
#44 | 5078 | Barrie Slaymaker | - VCP::Source::parse_time() 0s out undefined/missing fields | ||
#43 | 4500 | Barrie Slaymaker | - Minor POD cleanup | ||
#42 | 4497 | Barrie Slaymaker |
- --rev-root documented - All destinations handle rev_root defaulting now |
||
#41 | 4487 | Barrie Slaymaker | - dead code removal (thanks to clkao's coverage report) | ||
#40 | 4135 | Barrie Slaymaker | - Time fields may have trailing AM/PM or A/P without leading whitespace | ||
#39 | 4134 | Barrie Slaymaker |
- "AM", "PM", "A", and "P" (case insensitive) are now parsed properly when parsing time values |
||
#38 | 4039 | Barrie Slaymaker |
- VCP::Source::scan_metadata() API now in place, - VCP::Source::copy_revs() is fully deprecated. |
||
#37 | 4021 | Barrie Slaymaker |
- Remove all phashes and all base & fields pragmas - Work around SWASHGET error |
||
#36 | 3982 | Barrie Slaymaker |
- VCP::Source no longer leaks memory by delete()ing from a phash - VCP::Source::cvs now flushes to disk more often to conserve RAM |
||
#35 | 3970 | Barrie Slaymaker |
- VCP::Source handles rev queing, uses disk to reduce RAM - Lots of other fixes |
||
#34 | 3922 | Barrie Slaymaker | - More paranoid paramter checking | ||
#33 | 3916 | Barrie Slaymaker | - Reduce memory consumption | ||
#32 | 3907 | Barrie Slaymaker | - Debugging cleanups | ||
#31 | 3898 | Barrie Slaymaker | - VCP::Source::* --rev-root reinstanted | ||
#30 | 3855 | Barrie Slaymaker |
- vcp scan, filter, transfer basically functional - Need more work in re: storage format, etc, but functional |
||
#29 | 3835 | Barrie Slaymaker |
- VCP::Source supports queuing of revs and facilities for sending revs ASAP to conserve memory |
||
#28 | 3820 | Barrie Slaymaker |
- VCP::Source::revml now uses VCP::Source's queueing methods - For maintainability only, does not decrease memory util. |
||
#27 | 3819 | Barrie Slaymaker | - Factor send & queueing of revs up in to VCP::Source | ||
#26 | 3811 | Barrie Slaymaker | - fetch_*() and get_rev() renamed get_source_file() | ||
#25 | 3806 | Barrie Slaymaker | - VCP::Source no longer tries to send to a missing dest | ||
#24 | 3804 | Barrie Slaymaker | - Refactored to prepare way for reducing memory footprint | ||
#23 | 3706 | Barrie Slaymaker | - VCP gives some indication of output progress (need more) | ||
#22 | 3687 | Barrie Slaymaker | - Destinations may now use compile_path_re() | ||
#21 | 3681 | Barrie Slaymaker | - VCP now scans much more of real_vss_1 and converts it to revml | ||
#20 | 3679 | Barrie Slaymaker | - VCP::Source::vss respects --case-sensitive in more places | ||
#19 | 3677 | Barrie Slaymaker |
- rev_root sanity check is now case insensitive on Win32 - Parens in source filespecs are now treated as regular characters, not capture groups - ** is not treated as '...' |
||
#18 | 3477 | Barrie Slaymaker | - Make --rev-root only available in VCP::Source::p4 | ||
#17 | 3460 | Barrie Slaymaker |
- Revamp Plugin/Source/Dest hierarchy to allow for reguritating options in to .vcp files |
||
#16 | 3445 | Barrie Slaymaker |
- Don't misparse YYYY/MM/DD dates as MMMM/DD/YY. - t/61sort.t no longer blows up due to VCP::Rev's new BUG checks. |
||
#15 | 3443 | Barrie Slaymaker |
- Use BUG instead of Carp::confess - Recognize MM/DD/YY format dates |
||
#14 | 3157 | Barrie Slaymaker | debug conversion to VCP::Logger | ||
#13 | 3155 | Barrie Slaymaker |
Convert to logging using VCP::Logger to reduce stdout/err spew. Simplify & speed up debugging quite a bit. Provide more verbose information in logs. Print to STDERR progress reports to keep users from wondering what's going on. Breaks test; halfway through upgrading run3() to an inline function for speed and for VCP specific features. |
||
#12 | 3133 | Barrie Slaymaker |
Make destinations call back to sources to check out files to simplify the architecture (is_metadata_only() no longer needed) and make it more optimizable (checkouts can be batched). |
||
#11 | 3131 | Barrie Slaymaker |
Double the speed of the RCS file parser. Deprecate VCP::Revs::shift() in favor of remove_all(). |
||
#10 | 2824 | John Fetkovich |
removed CVS_CONTINUE field from Source/cvs.pm, and added CONTINUE field and continue accessor to Source.pm. Moved parsing of the --continue option also. |
||
#9 | 2809 | Barrie Slaymaker |
Implement --repo-id in Plugin.pm, refactor source & dest options parsing starting in VCP::Source::cvs (need to roll out to other sources and dests), get t/91cvs2revml.t passing again (first time in months! branching and --continue support works in cvs->foo!). |
||
#8 | 2453 | John Fetkovich |
removed compilation of revml. will be making that a separate executable. |
||
#7 | 2293 | Barrie Slaymaker | Update CHANGES, TODO, improve .vcp files, add --init-cvs | ||
#6 | 2015 | Barrie Slaymaker | submit changes | ||
#5 | 1998 | Barrie Slaymaker | Initial, revml and core VCP support for branches | ||
#4 | 1809 | Barrie Slaymaker | VCP::Patch should ignore lineends | ||
#3 | 628 | Barrie Slaymaker | Cleaned up POD in bin/vcp, added BSD-style license. | ||
#2 | 468 | Barrie Slaymaker |
- VCP::Dest::p4 now does change number aggregation based on the comment field changing or whenever a new revision of a file with unsubmitted changes shows up on the input stream. Since revisions of files are normally sorted in time order, this should work in a number of cases. I'm sure we'll need to generalize it, perhaps with a time thresholding function. - t/90cvs.t now tests cvs->p4 replication. - VCP::Dest::p4 now doesn't try to `p4 submit` when no changes are pending. - VCP::Rev now prevents the same label from being applied twice to a revision. This was occuring because the "r_1"-style label that gets added to a target revision by VCP::Dest::p4 could duplicate a label "r_1" that happened to already be on a revision. - Added t/00rev.t, the beginnings of a test suite for VCP::Rev. - Tweaked bin/gentrevml to comment revisions with their change number instead of using a unique comment for every revision for non-p4 t/test-*-in-0.revml files. This was necessary to test cvs->p4 functionality. |
||
#1 | 467 | Barrie Slaymaker | Version 0.01, initial checkin in perforce public depot. |