Revision history for Constant::Export::Lazy

0.19 2016-09-24 07:24:48

    - Updated Dist::Zilla::Plugin::ReadmeFromPod to fix an encoding
      issue that rendered by name as "�var Arnfj�r� Bjarmason"
      in the generated README.

0.18 2016-09-22 09:53:54

    - Only a documentation & test changes. Documented that if the
      "override" callback always provides a value you can omit the
      "call" callback, which was always implicitly supported but not
      documented or tested.

0.17 2016-07-15 21:10:49

   - Fix grammar & POD formatting in POD.

0.16 2016-07-14 15:29:51

   - My changes in 0.15 broke things for anyone using this package in
     conjunction with UNIVERSAL. See the end of the section discussing
     wrap_existing_import in the documentation for more details.

0.15 2016-06-01 13:42:03

   - Continue the change I made in 0.14 and replace the remaining
     occurrences of manual CODE symbol table munging with ->can()
     where possible.

0.14 2015-12-09 13:40:03

   - Simplify a lot of special cases around CODE / constant.pm
     generated constants by using more generic syntax to call our
     constants.

     We now just use ->can($name)->() instead of all this manual
     munging of the symbol table.

0.13 2014-12-27 21:23:06

    - Fix a long-standing bug of being unable to define constants from
      sub-packages at "call" time.

      If you were in the callback for YourExporter::CONSTANT and
      required a module that also used YourExporter module to get some
      YourExporter::UNRELATED_CONSTANT we wouldn't create that symbol
      for that module and we'd most likely die at compile-time.

      The reason for this is that we had a really naïve notion of when
      we should be interning constants purely based on whether we were
      in a nested "call" subroutine. Now we keep track of what
      constants packages want and intern those symbols they request.

    - We now use FATAL "recursion" warnings in Constant::Export::Lazy,
      if we've gone above perl's built-in recursion limit of 100 we'd
      rather die early than slowly exhaust all the memory on the
      system and have the OOM killer kill us when this happens.

    - Get rid of the Constant::Export::Lazy::Ctx::new()
      method. Instead we just bless directly into the package, this is
      just a nano-optimization.

    - Add a new test to see if we're actually defining constants, this
      test piggy-backs on the warnings emitted when our constant
      subroutines are overridden.

    - Clarify & reword some POD & inline comments.

0.12 2014-12-19 09:44:37

    - Added COPYRIGHT AND LICENSE section to pod, with standard "as perl"
      boilerplate.

    - Specified min perl version 5.6.0 in both code and metadata.

    - I don't know if this software actually works in 5.6.0, but
      you're welcome to try. All changes in this release brought to
      you by Neil Bowers via
      https://github.com/avar/constant-export-lazy/pull/2.

0.11 2014-12-01 21:03:29

    - This release only has changes to make the test suite more
      portable, nothing else.

    - Solve RT #95756 with the help of SPROUT, now we use B:: to test
      if we actually generated constants instead of using
      Data::Dump::Streamer.

    - We're also more forgiving about failures in our B::Deparse
      tests, and skip those tests on platforms where shelling out to
      -MO=Deparse doesn't work.

    - We now have no non-core test Prereqs. We just need a stock
      v5.8.0 to have all the prereqs we need, and in theory this
      module should work & run the test suite as far back as 5.000
      with the right test prereqs installed.

0.10 2014-11-05 10:44:33

    - The symbol exporter we were using was too eager and would export
      the entire *GLOB of a given constant we were exporting. So as an
      unintended side-effect exporting the constant CONST would also
      defined $FOO, @FOO, %FOO etc. in that package. Now instead we'll
      only export &FOO.

      Upgrading to this release could cause compile-time errors in
      existing code that implicitly relies on Constant::Export::Lazy
      to define these package-level symbols, but those cases should
      reveal genuine bugs. Code that doesn't rely on that side-effect
      should work as before.

0.09 2014-11-01 18:24:11

    - Attempt to future-proof how we define the constant subs for
      possible future perl core changes, see the "Is it time to
      separate pad names from SVs?" thread on perl5-porters. No
      functional changes for currently released perls.

    - Minor POD fixes from a couple of patches sent in via RT #99956
      and pull request #1 on GitHub.

0.08 2014-08-29 10:57:41

    - Improve the error message we show when we can't find a constant,
      now actually mention the constant we can't find.

    - This module actually worked on 5.8, but I recently broke 5.8
      support by adding a test that used 5.10 features, that's now
      fixed and we'll skip that test on <5.10 (it's just for testing
      the synopsis, not core functionality).

0.07 2014-05-06 21:51:39

    - No code changes at all. Just added tests to show that defining
      constant subs like I'm doing it doesn't work under 5.19.x (see
      "Re: How about having a recommended way to add constant subs
      dynamically?" on perl5-porters) --
      http://code.activestate.com/lists/perl5-porters/206929/

      I'll fix that in future releases, but first I want to get some
      test coverage on how defining constants works (or doesn't work)
      on different perl versions.

0.06 2014-03-10 21:14:32

    - Test & documentation for the API of private_name_munger

    - Trivial POD formatting error fix, no code changes at all.

    - Other POD rewording/documentation fixes / additions.

0.05 2014-02-23 16:37:00

    - Fix a minor failing test failing due to making an assumption
      about the internal workings of constant.pm that worked in 5.19.*
      but not in earlier perl versions.

0.04 2014-02-22 14:28:35

    - This is a major release

    - Add support for a buildargs callback. This can be used to add
      support for export tags, or to map the parameters to import() in
      any other way you can think of.

    - A new private_name_munger option is now available both globally
      and locally per-constant. This can be used to obscure the
      internal names of generated constants, to avoid hard to track
      down errors related to using lazy constants someone else might
      define for you (and then suddenly stop doing so).

    - When using the wrap_existing_import option the $ctx->call()
      method wouldn't work with some constant.pm-generated
      subroutines. This has now been exhaustively fixed.

    - After a lot of playing around with test coverage the test
      coverage of this module is now 100%. A few minor bugs were fixed
      in the process of achieving this.

    - Some minor documentation improvements, mainly rewording existing
      sentences.

0.03 2013-09-15 15:36:42

    - Fix trivial POD errors I didn't spot in the 0.02 release.

0.02 2013-09-15 15:28:10

    - Huge release with lots of new features and bugfixes.

    - Added an 'after' callback.

    - Added a 'stash' method / option.

    - Don't re-generate constants if someone subclasses the consatnt
      exporter.

    - Only export constant when explicitly requested, not everything
      we $ctx->call(...). Removes leaks in the API.

    - Correct behavior for skipping the interning of constants in the
      stash during override *only* for those constants we're
      overriding, not everything else.

    - Add tests for interop with constant.pm

    - Huge POD rewrite and lots of tests / examples added.

0.01 2013-08-30 12:32:23

    - Initial version