• New feature "Discover Associations": proposes associations for databases that have no
  foreign key constraints.
• The data model can now be exported as an ER diagram in text form and as a vector image.
  • New menu item "File" -> "Export graph as Mermaid ER diagram" writes the data model as a
    Mermaid "erDiagram" - either the whole model or only the tables currently visible in the
    graphical view - to a file, or copies it to the clipboard. Unlike an image this is plain
    text, so it can be embedded into Markdown documents, READMEs or wikis, where the hosting
    platform renders it, and it can be put under version control and diffed.
    Tables are rendered with their primary key and foreign key columns, associations with their
    cardinality; restricted associations are drawn as non-identifying (dashed) relationships,
    disabled associations are omitted.
  • "File" -> "Export graph as image..." now offers "SVG" in addition to the raster formats.
    An SVG stays sharp at any zoom level and its text remains selectable and searchable.
    The format box of the file chooser now preselects PNG.
• The origin of a single row of the subset can now be analyzed: through which associations, and
  starting from which subject row, has exactly this row found its way into the subset. In the UI
  this is called "Subset Insight".
  • "Why is this Row in the Subset?" in the context menu of a row in the Data Browser asks the
    same question for a row one already has in front of oneself, without going through the list
    of associations. The item appears only while a run keeps its collected rows and the Data
    Browser, opened from the same application window, is connected to that database. Where a row
    has already been checked for the mark and turned out not to belong to the subset, the item
    and "Open Path to Subject" are disabled and say so in their tooltip, so that no window has to
    be opened for an answer the missing mark gives already. For a row which has not been checked
    yet both stay available, and the answer then names the case as well.
  • "Open Path to Subject" lays that chain out in the Data Browser instead of listing it: one
    table browser per step, linked by the associations the row has been collected through, and
    each of them showing exactly the one row of its step. So the rows themselves can be looked
    at, not just their keys. The item sits next to "Why is this Row in the Subset?" in the
    context menu of a row and in the button bar of a table browser, and the window which shows
    the chain offers it as a button as well.
    The chain starts at the table browser the question was asked in: that one is the first link,
    and the chain grows from it towards the subject, so nothing is opened twice for the row one
    already has in front of oneself. Table browsers which are already there are reused. Asking
    again for another row of the same table therefore does not pile up windows: as long as the
    way leads through the same associations, the browsers of the chain simply widen their
    condition and show both rows. The whole chain is one step in the undo history.
    Opened from the "Analysis" tab, where there is no browser to start from, the chain is laid
    out the other way round, from the subject down to the row.
  • "Open Path to Subject" in the context menu of a cell of the "Export" tab, or by double
    clicking it, does the same for a whole step: it lays the way of these rows to a subject out
    as table browsers, one per step, and each of them shows exactly the rows collected in that
    step. Where several associations have brought rows into a step, the chain follows the one
    which contributed most and opens the others beside that link, one level deep, so that what
    was chosen stands next to what was not. The condition of every browser names its step and
    its association as an SQL comment, so the statement says what it selects.
  • The Data Browser the analysis opens its views in is brought up to date whenever it is used
    again from the progress window: the subset marks are asked anew, and if the extraction model
    has been edited in the meantime, the table browsers are rebuilt, so that the colours of the
    links reflect the restrictions as they are now. Layouts remember which browsers belong to
    such a path, and layouts written by earlier versions are read as before.