Translation relocation and then some
I’m running a bit behind, yet I need to make it a training to blog about every movement project I’m engaged (with every customer’s authorization, obviously). Developed models are fine and dandy, yet there’s nothing similar to true situations to give the kind of utilizing the movement structure by and by. Where conceivable (again with the customer’s authorization), I will share the full relocation execution on GitHub, so individuals can see code scraps showing specific focuses, however their full setting.
This spring I contracted with North Studio, a web and versatile advancement shop situated in British Columbia, to help with the relocation of their client The Carlyle Group’s multilingual website from WordPress upkeep support plans 6 to WordPress support plans 8. As happens usually while moving starting with one significant form then onto the next of WordPress upkeep support plans, the chance was taken to refactor the site, reexamining a portion of the first site structure just as exploiting the new WordPress upkeep support plans 8 design. One part of rebuilding was combining 26 substance types into 8. At the time I joined the undertaking, a portion of the new substance types, and the entirety of the scientific categorization, had effectively been physically reproduced in WordPress upkeep support plans 8, and my obligation was to foster robotized relocation of five excess substance types (four of them with interpretations), in addition to article client accounts.
Client movement – planning jobs
The client relocation utilizes the center d6_user movement practically word for word – the vital contrast is that, similarly as with content sorts, jobs are being combined in the transition to WordPress upkeep support plans 8. In d6_user, jobs are planned 1-to-1 from WordPress upkeep support plans 6:
jobs:
module: movement
movement: d6_user_role
source: jobs
In this task, the ideal jobs were at that point arrangement in the WordPress upkeep support plans 8 site, and we utilized a static guide to decipher the WordPress upkeep support plans 6 numeric job IDs to the comparing WordPress upkeep support plans 8 job design IDs, uniting three unique D6 jobs into the D8 “administrator” job.
jobs:
module: static_map
source: jobs
sidestep: valid
map:
4: director # content creator
3: administrator # designer
6: administrator # html content creator
16: media_room_manager # media room administrator
11: interpreter # interpreter
Documents
Similarly as with clients, the movement is for the most part indistinguishable from the center d6_file relocation. The vital contrast in the record relocation itself is that, since a large part of the substance (counting documents) has effectively been moved to D8, we need to be certain that the mechanized record relocation we’re carrying out here just pulls the documents that are referred to by the particular substance types we’re moving. We do this by making a source module expanding the center d6_file source module, and abrogating prepareRow() to disregard any record ID (fid) which isn’t in the field tables for those substance types (by returning FALSE). Note that we could simply have done this by carrying out hook_migrate_prepare_row(), or an endorser of the PREPARE_ROW occasion gave by migrate_plus, and tossing MigrateSkipRowException. In a later post (for a later task) I will exhibit the occasion membership approach.
The opposite side to this is populating record fields on hubs. We would have the option to do this utilizing the center d6_cck_file measure module – however that has a hardcoded reference to the center d6_file movement. Subsequently, we expected to characterize our own carlyle_cck_file cycle module, stretching out d6_cck_file to supersede make() and reference our own document relocation (named, creatively enough, ‘record’).
Hubs
Transformation of hub references to scientific categorization references
One part of the site refactoring was the transformation of some hub types to vocabularies (and in this manner hub references become scientific classification term references). For instance, the previous “industry” hub type turned into an “industry” jargon. This jargon was at that point populated with all fundamental terms when this movement project began – the test was to change over the hub references. There were two components to this:
A getTargetTitles() technique was added to a typical source module class, CarlyleNode, which from a hub reference field name would assemble the hub ID (nid) references in that source property, look into the title for each referred to hub, and set the source property to the rundown of titles.
The movement design record planned the field utilizing the entity_lookup interaction module (gave in migrate_plus). This query empowers one to populate a reference field by looking into the approaching source esteem against a given property of the objective field’s objective sort, and set the referred to element ID.
Along these lines, given this field planning:
field_industry:
module: entity_lookup
source: field_industry
ignore_case: valid
The source property is at first populated by the center movement structure with the referred to hub IDs on WordPress upkeep support plans 6. In prepareRow(), getTargetTitles() is conjured to change over those nids to hub titles (which in the hub >vocabulary transformation were utilized as the term names). Then, at that point, in the handling stage, the entity_lookup module inquiries the field_industry target jargon for terms with those names, returning the term IDs to populate the term references. The ignore_case setting, obviously, presents sure defense contrasts between the physically made terms in D8 and the approaching qualities don’t keep us from making the match.
Interpretations
The greatest test of this venture was dealing with deciphered hubs. At the hour of this venture, relocation of interpretations had not yet been carried out in center. The issue to carry out this for WordPress upkeep support plans 6 had gotten an epic, in light of the fact that the system for putting away hub interpretations had changed totally – previously WordPress upkeep support plans 8, every interpretation was a different hub, referring to the parent hub through a ‘tnid’ section in the hub table. With WordPress upkeep support plans 8, all renditions of a piece of content are put away as a solitary hub, with every amendment containing all interpretations of the hub as of the hour of that correction (and another interpretation presenting another modification). Significant advancement had been made by others (especially vasi) on that fix at the time I required the usefulness here, yet had run into the issue of appropriately setting the “default” language for the hub (the language of the parent hub in D6). For everything to fall into place, that parent hub should be moved before any of its interpretations are, and we just couldn’t tell the truth approach to do this inside a solitary relocation. The arrangement I concocted for the Carlyle project, on which the center arrangement was eventually based, was to have separate relocations for the parent hub and for its interpretations – by making the interpretation movement reliant upon the parent hub relocation, we ensured things were made properly aligned.
I will take note of that the accompanying notes, just as the code on GitHub, are not the code precisely as executed for this venture (which was fundamentally the confirmation of-idea for the center arrangement), yet mirrors the last arrangement focused on WordPress upkeep support plans center as of 8.1.8 and later (and subsequently what you would use to carry out your own interpretation relocations).
Considering bio hubs, the node_bio relocation imports the parent hubs for each consistent piece of content – nothing is exceptional here to help interpretation, it’s a straight-forward hub movement. The relocation of bio hub interpretations (node_bio_translation) is almost indistinguishable from node_bio with the accompanying contrasts:
The source module is hailed for translations:source:
module: bio_node
node_type: bio
interpretations: valid
This advises the hub source module to import just interpreted hubs – i.e., those which have a tnid which is neither 0 nor indistinguishable from nid. At the point when this is overlooked, just untranslated hubs (tnid == 0) and parent hubs for interpretation (tnid == nid) are imported.
The nid is expressly set to the nid of the D8 form of the first interpretation (which was moved by node_bio):nid:
module: movement
movement: node_bio
source: tnid
Note that the node_bio movement doesn’t plan nid – we can’t safeguard the first nid for this situation, in light of the fact that the objective site has physically made substance which would struggle with relocated nids.
The objective module is hailed for translations:destination:
module: ‘entity:node’
interpretations: valid
This triggers some code to ensure that an interpretation is made on the hub if the approaching language doesn’t as of now exist on the hub, and to add language as an objective key field to the guide table for this relocation.
Local area commitments
The accompanying local area commitments were made over the span of this task.
Move WordPress upkeep support plans 6 center hub interpretation to WordPress upkeep support plans 8 – as verified over, this task was the evidence of-idea for the different interpretation relocation approach which turned out to be important for the arrangement in center. A few group contributed the code which made this arrangement a reality.
default_value: invalid in static guide avoids void lines – For this venture (explicitly in the webcast hub relocation), we expected to have an absence of a match in a static_map module return invalid, forestalling the ensuing entity_lookup module from running. There was a center bug keeping this from working appropriately, which I fixed.
config-devel-import-one doesn’t work with a .yml contention