String segments resests after synchronizing translations

[Azadron228]

I am encountering an issue where some string segments reset after synchronizing and publishing translations. Specifically:

Only certain pages are affected, and within those pages, only some strings reset.
The resetting string segments are being re-inserted into the wagtail_localize_stringsegments table after synchronization.
Other pages and string segments of same page model work as expected without issues.

Environment:

wagtail-localize: 1.8
Django: 4.0.7
Wagtail: 5.2
Database: MySQL

This is how i created page

class ChairMemberPage(SitemapMixin, Page):
    parent_page_types = [ChairPage]

    photo = models.ForeignKey(
        "wagtailimages.Image",
        null=True,
        on_delete=models.SET_NULL,
        related_name="+",
        verbose_name="Фото",
    )
    position = models.CharField(max_length=100, verbose_name="Должность")
    email = models.EmailField(verbose_name="Электронная почта")
    subjects = RichTextField(verbose_name="Основные преподаваемые дисциплины")
    education = RichTextField(verbose_name="Образование")

    info = StreamField(
        [("paragraph", blocks.RichTextBlock(label="Параграф с доп. информацией"))],
        verbose_name="Дополнительная информация",
        blank=True,
        null=True,
        use_json_field=True,
    )

    content_panels = Page.content_panels + [
        FieldPanel("photo"),
        FieldPanel("position"),
        FieldPanel("email"),
        FieldPanel("subjects"),
        FieldPanel("education"),
        FieldPanel("info"),
    ]

    search_fields = Page.search_fields + [
        index.SearchField("position"),
        index.SearchField("email"),
        index.SearchField("subjects"),
        index.SearchField("education"),
        index.SearchField("info"),
    ]

    class Meta:
        verbose_name = "Сотрудник кафедры"
        verbose_name_plural = "Сотрудники кафедры"

[Azadron228]

After further analysis, I identified the root cause of a data inconsistency issue during the synchronization process. The issue occurs because wagtail_localize_stringsegment entries are being reinserted instead of being retrieved from the database, and not all segments are inserted as expected. This leads to a mismatch between wagtail_localize_stringtranslation.translation_of_id and wagtail_localize_stringsegment.string_id.

This query run for almost all segments

INSERT INTO `wagtail_localize_stringsegment` (`source_id`, `context_id`, `order`, `string_id`, `attrs`) 
VALUES (1597, 16609, 52, 29431, 'null');

SQL Query to Retrieve Translations:

The following query retrieves translations for string segments:

SELECT
    `wagtail_localize_stringsegment`.`id`,
    `wagtail_localize_stringsegment`.`source_id`,
    `wagtail_localize_stringsegment`.`context_id`,
    `wagtail_localize_stringsegment`.`order`,
    `wagtail_localize_stringsegment`.`string_id`,
    `wagtail_localize_stringsegment`.`attrs`,
    (
        SELECT
            U0.`data`
        FROM
            `wagtail_localize_stringtranslation` U0
        WHERE
            U0.`context_id` = `wagtail_localize_stringsegment`.`context_id`
            AND U0.`locale_id` = 2
            AND U0.`translation_of_id` = `wagtail_localize_stringsegment`.`string_id` --at this line filtering fails
            AND NOT U0.`has_error`
    ) AS `translation`,
    `wagtail_localize_translationcontext`.`id`,
    `wagtail_localize_translationcontext`.`object_id`,
    `wagtail_localize_translationcontext`.`path_id`,
    `wagtail_localize_translationcontext`.`path`,
    `wagtail_localize_translationcontext`.`field_path`,
    `wagtail_localize_string`.`id`,
    `wagtail_localize_string`.`locale_id`,
    `wagtail_localize_string`.`data_hash`,
    `wagtail_localize_string`.`data`
FROM
    `wagtail_localize_stringsegment`
INNER JOIN `wagtail_localize_translationcontext`
    ON `wagtail_localize_stringsegment`.`context_id` = `wagtail_localize_translationcontext`.`id`
INNER JOIN `wagtail_localize_string`
    ON `wagtail_localize_stringsegment`.`string_id` = `wagtail_localize_string`.`id`
WHERE
    `wagtail_localize_stringsegment`.`source_id` = 1597;

[Azadron228]

Wrong segment comes from extract_segments() which leads to data inconsistentcy, and recreation of string.

I have checked strings in database got two visually similar strings which differ in backspace which lead to bug.