- Koha-devel - lists.koha-community.org

Resultsets vs. list context
by Tomas Cohen Arazi 10 Jan '22

10 Jan '22

Hi, let me tell you a short and (technical but) interesting story. Most of you might know about this, but I found it could be useful to know and/or to refresh. DBIC has a nice property that our Koha::Objects inherited (obviously): you can chain calls like this: Koha::Patrons->search( $filter_1 )->search( $filter_2 )->search( $filter_3 )->...; We recently agreed (not sure there is a coding guideline) that if we wanted to provide high-level methods to encapsulate such filters, we would call those methods 'filter_by_*' unless the context suggests a better name. So we could, in theory do something like: my $cool_old_books = Koha::Biblios->search( $some_initial_filter ) ->filter_by_older_than({ years => 30 }) ->filter_by_coolness(); So far, so good :-D But then, we have another (cool) method in Koha::Objects: _new_from_dbic. This method allows us to, given a DBIC relationship, convert the linked resultset into the proper Koha::Object(s)-based class. For example: my $patron_checkouts = $patron->checkouts; in this case, the actual code is pretty simple thanks to _new_from_dbic: sub checkouts { my ($self) = @_; my $checkouts = $self->_result->issues; return Koha::Checkouts->_new_from_dbic( $checkouts ); } Methods like ->search behave consistently to what DBIC does, and as such, they honor list context. This is: my $scalar = Koha::Patrons->search(filter); my @list = Koha::Patrons->search($filter); will do what you think: $scalar will be an object, representing a resultset, and @list will be the actually fetched list of results from the search, as a list. But there's a catch. And we need to be aware of that and there is an ongoing discussion about it: _new_from_dbic doesn't honor list context [1]. This means when you code, you need to go look if the method you're calling is _new_from_dbic-based, or ->search based [2]. This is not cool. @checkouts = $patron->checkouts; # doesn't do what you expect! @checkouts = $patron->checkouts->filter_by_overdue; # this does if ->search based. Our codebase is full of places in which we either call ->as_list in our resultsets (because something exploded if we didn't, for example): my @checkouts = $patron->checkouts->as_list; or in which we explicitly call `scalar the_thing' so we are sure things are evaluated in scalar context because we want the resultset instead (for example, when passing params to a template): $template->param( checkouts => scalar $patron->checkouts ); I proposed we make ->_new_from_dbic and ->empty honor list context and hence making our methods consistent (can be seen by the referenced bugs). But we found a nasty problem: calling methods in Template::Toolkit internally assigns them to a global stash which is implemented as a hash (a.k.a. list context :-D). So, you will find that 'some' things are calculated on the controller (.pl) scripts and then passed to the templates (to overcome this), and some others are calculated directly in the templates. Now you know the reason we've introduced those things over the years in the codebase. So, making _new_from_dbic honor list context, opens a big can of worms in the template front. We explored the idea of using a different Template::Toolkit stash, that honors scalar context, but the author has marked it as experimental, and it would still require changing many things in the templates. The tradeoff doesn't lean towards that path. The conclusion so far is that we should drop support for list context. Hence, my suggestion as developer and QA team member is: TL;DR Make use of ->as_list when you have a resultset really need list context like in: foreach my $item ( @{$items->as_list} ) { ... } grep { filter($_) } @{@libraries->as_list} my @biblios = $biblios->as_list; and so on. This will save you some headaches (in some cases) and will make your code survive this upcoming change. We should start enforcing this in the QA step as well, so adjust your code for this. Once we remove wantarray, things will explode on their own. Best regards [1] https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=28883 [2] There is also Koha::Objects->empty, actually. That presents the same problem as _new_from_dbic. https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=28871 -- Tomás Cohen Arazi Theke Solutions (http://theke.io) ✆ +54 9351 3513384 GPG: B2F3C15F

2 1

New class: Koha::Result::Boolean
by Tomas Cohen Arazi 07 Jan '22

07 Jan '22

Hi all, happy 2022 :-D I wanted to mention that a bug that introduces Koha::Result::Boolean has been pushed [1]. This class provides a way to return an object, that evaluates to a boolean in (Perl's) bool context but also carries reasons for the result. It has been used in a couple places, like bug 29765 [2] and bug 29788 [3] (which still needs the QA step). [1] https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=29746 [2] https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=29765 [3] https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=29788 -- Tomás Cohen Arazi Theke Solutions (http://theke.io) ✆ +54 9351 3513384 GPG: B2F3C15F

2 1

Dev meeting 5 January 2022 report
by Fridolin SOMERS 06 Jan '22

06 Jan '22

Hi, Here is Dev meeting 5 January 2022 report : https://wiki.koha-community.org/wiki/Development_IRC_meeting_5_January_2022 Next dev meeting : https://wiki.koha-community.org/wiki/Development_IRC_meeting_19_January_2022 There is also a general meeting soon : https://wiki.koha-community.org/wiki/General_IRC_meeting_12_January_2022 See you around :D Best regards, -- Fridolin SOMERS <fridolin.somers(a)biblibre.com> Software and system maintainer 🦄 BibLibre, France

3 2

Koha 19.11.24 released
by Wainui Witika-Park 05 Jan '22

05 Jan '22

Hello everyone! The Koha community is proud to announce the release of version 19.11.24. The full release notes can be found at: https://koha-community.org/koha-19-11-24-released/ Thank you to everyone involved! Cheers, Wainui Witika-Park (she/her)

1 0

Call for volunteers to work on ElasticSearch 7 support (ES6 will be EOL on February 8)
by Victor Grousset/tuxayo 02 Jan '22

02 Jan '22

Hi :) After ES6's EOL, if another vulnerability if found in Log4J for example, Koha instances using ES won't be able to patch it. Here is the omnibus ticket: https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=25439 There are *at least* a 1/3 of them that look quite easy. Cheers, -- Victor Grousset/tuxayo

1 0

KohaCon22 proposals changes suggestions
by Fridolin SOMERS 01 Jan '22

01 Jan '22

Hi, We are getting close to end date for KohaCon22 proposals, January 8. Unfortunately there is actually no proposal : https://wiki.koha-community.org/wiki/KohaCon22_Proposals This as been discussed in last dev meeting. https://wiki.koha-community.org/wiki/Development_IRC_meeting_22_December_20… The world situation on Covid-19 virus is not helping, we hope it will be better soon. First we suggest to extend end date to mid-February. Second we suggest to allow proposals outside eligible continents, keeping priority to actual ones. We may have full-online proposals, we will see. If no oppositions on those changes, I will announce this beginning January. Best regards, -- Fridolin SOMERS <fridolin.somers(a)biblibre.com> Software and system maintainer 🦄 BibLibre, France

3 2

Koha packages are available 🎁
by Mason James 30 Dec '21

30 Dec '21

kia ora community, the latest Koha packages are available cheers, Mason

4 6

Development IRC meeting 5 January 2022
by Fridolin SOMERS 29 Dec '21

29 Dec '21

Hi See you at next dev meeting : https://wiki.koha-community.org/wiki/Development_IRC_meeting_5_January_2022 The calendar event is not yet created, will be soon. Best regards, -- Fridolin SOMERS <fridolin.somers(a)biblibre.com> Software and system maintainer 🦄 BibLibre, France

1 0

API error responses change
by Tomas Cohen Arazi 27 Dec '21

27 Dec '21

Hi all, I wanted to mention that I've submitted patches for bug 28020 [1]. The change introduces a new attribute to the generic error response object that is reused everywhere in the spec. The new attribute is called 'error_code' and is intended to be adopted so API consumers have better means to deal with error responses. Remember that right now we return an error status (e.g. 409) and a (too) generic description in textual form (e.g. Conflict found). A good use case for this can be seen here [2], in which we already have a handy error string, but we translate it to some text, while we should be returning the error_code attribute, and have it documented on the spec as proposed on bug 28020. Early feedback is welcome, as I would base several things on top of this! THANKS! [1] https://bugs.koha-community.org/bugzilla3/show_bug.cgi?id=28020 [2] https://bugs.koha-community.org/bugzilla3/attachment.cgi?id=128816 -- Tomás Cohen Arazi Theke Solutions (http://theke.io) ✆ +54 9351 3513384 GPG: B2F3C15F

1 0

Koha Community Newsletter: December 2021
by Koha Newsletter 26 Dec '21

26 Dec '21

The Koha Community Newsletter for December 2021 is here: https://koha-community.org/koha-community-newsletter-december-2021/ Many thanks to everyone who submitted articles and news to this newsletter! Please feel free to email me with any corrections or suggestions. -- Michael Kuhn Editor, Koha Community Newsletter

1 0