A5 Tree check

What the app watches in a tree, with what thresholds, and what can be done with a finding: fix it, mark it Not an error, or let the research assess it. Describes the Strom app 3.21.2 (and the research 1.25.2). The research has its own check strom check (R5). It is not described here, except where the two agree.

A5.1 The check and its findings

The app checks the whole tree every time the check is opened. Nothing is saved and the tree does not change. Each finding has a severity:

• Limits:
  • Findings of the same kind form a group, sorted by severity, size and name. Only a group with an error is open. Over 50 findings, more are shown 20 at a time. On a phone (up to 640 px) Leads are collapsed when there is the section To check too.
  • Placeholder persons are not reported. A finding marked Not an error (A5.7) is neither counted nor shown.
  • A read-only tree only shows findings. For a tree other than the open one, a fix or a mark opens the tree first.
  • From app 3.27.0: the severity formerly called "warning" is called to check, in summaries and counts too ("2 errors, 1 to check, 3 leads."). The Table view shows findings right in the cells, and a finding marked Not an error as unusual (A6.7.5).
  • From app 3.27.0: the section of errors and findings to check is called To check (formerly Problems). A tree without them has "Nothing to check", with leads "Nothing to check. 3 leads."
• Controls: the dialog Tree validation: Tree overview → Data quality → Quick actions → Validation details, the row "n leads ›" in Consistency (opens Leads directly), the card with errors in Where to focus. From app 3.27.0 also Data quality › in a finding's bubble in the Table view: it opens the dialog at that finding's group (for a finding marked Not an error at the list of marked findings). After a GEDCOM import the app asks ("Data check") only when it finds errors or findings to check (A2). The same after a merge with another tree, for new findings. The check has no menu item of its own.
• Since: app 1.0.0; the sections Problems and Leads app 3.17.0; the name to check and the link from the Table view app 3.27.0; the section To check and "Nothing to check" app 3.27.0

A5.1.1 One-click fix

Some findings have a button Fix; the summary above the findings has Fix all (n) for all findings that can be fixed. A fix is one step back ("fix from validation") and is written into the change history.

• Under Fix all (n) stands what is fixed, with counts, only for the kinds the tree has: "Fixes: broken links between people and unions (2); a union of a person with themselves – removed (1); a duplicate union – merged into the first, children kept (1); a place written several ways – unified to the most used spelling (3). Other findings stay unchanged." In a read-only tree neither the button nor the line is there.

• Since: app 1.1.0; fixing places 1.10.0; the line under Fix all (n) app 3.27.0

A5.2 Family links

The check looks for links that cannot hold. All are errors, except a duplicate partnership (to check).

• Since: app 1.0.0; event participant 1.16.0

A5.3 Dates and age

The check reports only what certainly does not hold. It counts an incomplete or approximate date over its whole possible range: without a month January to December, "about" ± 2 years, "before" up to 20 years earlier, "after" up to 20 years later, a range up to its end. Only when the violation holds over the whole range is it a finding. The exception are the milder limits to check from app 3.27.0 (below): there it is enough when the limit is crossed by one end of the possible age.

• Limits: the research's strom check has the same exceptions after death. A parent without a type of relationship counts as biological.
• From app 3.27.0: milder limits to check:
  • a mother younger than 16 or older than 45 and a father younger than 18 or older than 60 at a child's birth. Only a biological parent of known sex, the age from both dates of birth;
  • a marriage at an age under 16;
  • it is enough when the limit is crossed by one end of the possible age (an age "15–16" is under 16). A date with about counts as an estimate of the age; a date with before or after is not compared;
  • only the stricter finding is reported: where Parent younger than child, Parent very young / very old or Married as a child already holds, the milder finding is not;
  • the recorded age (at a marriage the age of each of the couple, at death, at birth) is compared with the age from the date of birth and the date of the event. It is a finding when it is outside the possible range by more than 2 years. An age written in words or with about, before, after is not compared, nor when one of the dates is only an estimate;
  • the row of the finding shows the dates and the age ("* 1850 → * 1866: 15–16", "⚭ 1872: 30 ≠ 26–27").
• Since: app 1.0.0 (age of parents); 1.4.0 (events); 1.7.0 (dates and death); a child after the death of a parent of unknown sex 3.10.0; four exceptions after death 3.17.0 (before only burial); milder limits for the age of parents and of marriage and the recorded age 3.27.0

A5.4 Sources, attachments and merges on import

Older versions could give two people the same ID on import. The check reports it on a clear sign (more than 2 parents, a person who is their own parent, child or partner), or when a child is born less than 12 years after a parent and in addition there are parents or a partner of the same sex. Parents of the same sex alone do not give a finding. The remedy is in the text of the finding: import the original file again as a new tree.

• Since: app 1.7.0 (sources and attachments); 3.0.0 (photos); 3.9.0 (merges on import)

A5.5 Leads: godparents and spellings of places

Leads are not errors. They show where there may be kinship or an inconsistent spelling.

• A godparent who keeps turning up — often a relative: the same godparent or witness is at the events of at least 2 people who are not related to each other, and those people belong to at least 2 groups of siblings. The row shows at how many events and people. Example: Marie Vlková is godmother to the Horák children and to the children of their neighbours.
• One place written several ways: spellings that differ only in upper and lower case, diacritics, or a full stop or comma, e.g. "Brno" and "brno.". The row shows each spelling with a count; the fix is in A5.1.1.
• Limits: relatives are people at most 3 steps apart through parents, children and partners. Counted are the participants of a person's events, marriage witnesses and the participants of a couple's shared events. Leads also hold the age difference of partners (A5.3).
• Since: app 1.10.0 (places); 1.16.0 (godparents); groups of siblings and officials 3.17.0

A5.5.1 How the same godparent is recognised

• A person linked to the tree is recognised by the record. A name that is only written is compared as text; the row then has "matched by name".
• From the entry the name is taken up to the first comma, lower-case letter or digit. Titles and forms of address (godfather, Mr., Mgr. …) are left out, particles (von, de …) stay, text in brackets is a note. Comparison ignores diacritics and upper and lower case.
• The name must have at least 2 words. A given name alone never matches.
• Whoever is at the event by office is not reported: the officiant, the midwife and the clergy (parish priest, chaplain, Pfarrer, parochus …). A cleric recorded as a godparent is reported.

A5.6 Possible duplicate person

Two people may be the same when they have the same surname and the same given name and were born in the same years. Comparison ignores upper and lower case and diacritics. The finding is in Leads in the group Possible duplicate person.

• Limits:
  • Year of birth: the date of birth; without it the baptism (the year of baptism or the year before it); without it the death or burial with the age at death. When both years come from births, they must overlap. Otherwise they may be 2 years apart.
  • When one person lacks a given name, the surname and the years are enough.
  • People without a date are paired only with exactly the same name. A person with a date is not paired with a person without one.
  • Not reported: placeholder and unnamed persons, incompatible sex, parent and child, and pairs where one person certainly died before the other's birth (or baptism).
• Controls: Two different people (A5.7); at a verdict of the research Merge… (A5.6.1).
• Since: app 1.9.9; years from baptism and death 3.17.0

A5.6.1 The research's verdict

For a tree from the research, the app asks the research about all pairs when the dialog opens. The dialog does not wait for the answer. The research assesses only people who have its number (P…). The first rule that applies decides:

• Limits:
  • Pairs with the verdict Probably the same person and Merged in the research go to the top, then Uncertain, then those without a verdict.
  • Pairs Two people go aside into the row Told apart by the research: n and are not counted.
  • A verdict that arrives into an open group only adds a label. It is re-sorted at the next opening.
  • Under the group is The research is assessing… or Assessed by the research. When the research that assessed at the last connection is not running: "Not assessed by the research – the research is not running." A research that cannot assess is not mentioned.
• Controls: at a verdict Merge… (merging people) or Load the new version · terminal: strom person compare P… P…, for many pairs --pairs-from. Connection with the research: S1.
• Since: app 3.17.0 (research 1.22.0)

A5.6.2 Warning when entering a person

When a new person is entered, the app shows at most 3 similar people ("Similar people already exist:"). It compares differently from A5.6: by a similarity score as in merging, never with incompatible sex.

• Controls: Go to person, Use existing · turned off by Settings → Notifications → Duplicate suggestions.
• Since: app 1.4.0

A5.7 Not an error

A finding that is not an error is marked. Then it is no longer reported, even when the tree is checked again. The mark is a change of the tree: it has a step back and an entry in the change history. For a possible duplicate the button is called Two different people.

• Limits:
  • The app recognises the same finding by its kind, people and subject. The subject is the year for an event or wedding after death, the name for a godparent and the most frequent spelling for a place. When any of these changes, the finding shows again.
  • For a tree from the research, the mark is sent to the research when all people of the finding have a research number (at most 20 people). The research's GEDCOM file does not carry the marks; when a new version is loaded they are taken over from the old version of the tree.
  • The research respects marks at an event and a wedding after death: strom check does not report them and the agent does not investigate them.
  • State of a mark: Sent to the research, Waiting for the research (the research is not running, it is sent at the next connection), In the app only (the research did not take it or the people have no number; it is not sent again), Marked in the research.
  • A mark at a place has no people, so it is not sent to the research. When a part of the tree is separated, only marks whose people all stay are kept.
  • The app knows a mark written in the research only for an open tree from the research, when the research is running.
  • From app 3.27.0: the subject for Married under the age of 16 is the year of the marriage, for Recorded age differs by more than 2 years the fact (birth, marriage, death) and for a marriage its year. The Table view shows a marked finding as unusual, and from there it can be marked and restored too (A6.7.5).
• Controls: in the row of a finding Not an error / Two different people, for 6 s the offer Undo. The list Marked as not an error: n under the findings, at each mark a date or Marked in the research, the button Restore · terminal: strom quality list, strom quality dismiss possibleDuplicate P… P…, strom quality dismiss eventAfterDeath P… --subject 1861, strom quality undo …. Only a person writes a mark, never the agent.
• Since: app 3.17.0 (research 1.22.0)

A5.8 Tree overview → Data quality

The section sums up the check and the completeness of the tree. It counts the same way as the dialog, without marked findings and without pairs the research told apart.

• Limits:
  • Over 10 MB of photos and attachments a notice is added ("Media exceed 10 MB. The tree file saves and shares more slowly."). An orphaned place is a place nothing refers to.
  • An unconnected part has no path to the main family; it can be separated (Separate unconnected parts).
  • The card in Where to focus shows the number of errors and the most frequent kind. It has the button Fix (open tree) or Validation details; both open the dialog.
• Controls: Tree overview → Data quality; in the tree menu also Data quality directly. Actions that change data work only in the open tree.
• Since: app 3.13.0; leads and marks 3.17.0; "to check" instead of "warning" 3.27.0