Addon:QueryGramplet

From Gramps
Revision as of 01:53, 27 December 2013 by Dsblank (talk | contribs) (Examples)
Jump to: navigation, search

The Query Gramplet takes SQL-like queries and produces a Quick View.

Examples

The QueryGramplet in gramps-addons/trunk for gramps/master can now SELECT, UPDATE, and DELETE. Some examples (keywords are shown capitalized, but the SQL parser is case-insensitive):

DELETE FROM person WHERE given == "Travis";
SELECT * FROM person;
SELECT gramps_id, given, surname FROM person;
SELECT event_ref_list[0].ref FROM person;
UPDATE person SET given="Gary" WHERE given == "Travis"
SELECT gramps_id FROM person where ROWNUM < 10;
SELECT gramps_id FROM person LIMIT 5;
SELECT gramps_id FROM person LIMIT 20,30;

This API is made possible through the generic struct/json interface. It is very little code, because it relies on these generic structures. It should be able to be made solid enough to expose to users (say as a generic filter). The parser can be made more user friendly... it may just throw an error currently.

I'd be interested in any limitations you find, or enhancement ideas.

SQL

Here is the grammar for the subset of SQL supported. The SELECT, UPDATE, DELETE, and LIMIT clauses may be in any order. The WHERE clause (if used) must be last.

SELECT expr1 [as var1][, expr2 [as var2], ...] FROM table [LIMIT number1[, number2]] [WHERE expression];
UPDATE table SET field1=expr1[, field2=expr2, ...] [LIMIT number1[, number2]] [WHERE expression];
DELETE FROM table [LIMIT number1[, number2]] [WHERE expression];

table is one of:

  • person
  • place
  • repository
  • event
  • citation
  • source
  • tag
  • media
  • family

Other items:

  • expr is a field, *, or expression
  • var is an alias
  • number1 by itself is maximum number of rows to select
  • number1 with number2 is start, stop (first row is zero)
  • expression is any valid Python expression

expression and expr may use:

  • random.random() (or other random method)
  • ROWNUM (zero-based counter)
  • col[N] (alias to column)
  • aliases

The following shortcuts can be used in expressions and as a field:

  • surname, short for "primary_name.surname_list[0].surname"
  • given, short for "primary_name.first_name"

Notes

Some notes on use:

1) Most SQL clauses (UPDATE table, FROM table, SELECT ..., SET field=value, ..., LIMIT ...) can appear in any position, any order

2) ...except the WHERE clause: it must be last; this is because the WHERE clause is not parsed, because:

3) The WHERE clause uses any valid Python expressions. It imports some libraries (such as random), to have ready possible expression needed.

SELECT * from person WHERE random.random() < .1;

This selects records where each has a 10% chance of being selected.

4) The SELECT fields use the bracketed notation for list references. Use "event_ref_list[0]".

5) JOINS are not necessary, because it automatically looks up all relations through the handles.

6) UPDATE currently only works on the primary object, not on the joined object. For example, you can't update the birth date of an event through the person's referenced events. I think this can be fixed.

7) Tables are lowercase, single (not plural) form (eg, person, tag, event).

8) Implemented "LIMIT number", "LIMIT start, stop", and "WHERE ROWNUM < number" (ROWNUM can be used in any expression).

9) Field names are the actual names of the fields of the gramps.gen.lib objects, verbatim, no differences. You might need to look up what you need... no help yet from this interface (although I am working on defining a built-in Schema that could help)

10) .handle or .ref automatically look up their references.

11) Shortcut: you can use col[N] in the WHERE clause to reference a column selected.

SELECT gramps_id, private FROM person WHERE not col[1];

That will select all people where private is not True. That would include None (non-existent record) and False values. To select only False values, use:

SELECT gramps_id, private FROM person WHERE col[1] == False;

That will not select None values.

12) If an object doesn't match any selected field, it just doesn't show. For example, to find all of the people with at least two surnames on their primary name, use:

SELECT primary_name.surname_list[1] FROM person;

or

SELECT gramps_id, primary_name.surname_list[1] FROM person WHERE col[1];

13) In a SELECT (for speed reasons), you need to reference a field before you can use it. That is not necessary in the UPDATE or DELETE statements. (The idea here is that SELECTS are done quite frequently, but UPDATES are done rarely, and it doesn't matter if those take a little longer). Maybe we can relax this constraint.

14) The semicolon is optional

15) Be careful selecting all fields from all records... that could take up a lot of memory, and bring down gramps.

16) This should be fairly fast, but it does call eval(). This might make things a little slower, but made the code much easier to write. And it does use the full power of python.

17) You can use parens in an "UPDATE table SET field=value" value. Something like:

UPDATE table SET field=(field + 1);

but that hasn't been well-tested. (Speaking of testing, there is a Vassilii-inspired unittest with the QueryQuickview... will add more there).

18) The primary_name... stuff is really long and verbose. Maybe we need some "virtual columns".

19) Fields that contain other objects, or lists of objects, will show as dictionaries and lists of dictionaries. You can refine those fields by further specifying subparts. Maybe we should not show these, or show in another form...

20) If a selected field does not exist in a record, then it will have a value of None. For example, if you are selecting those people that have a second surname on their primary name, and there are some people who do not have a second surname, it will appear as None. If all columns are None, then the item will not be select at all.

Older documentation for the QueryGramplet in Gramps 3.4

This has different table names (people rather than person), and other names that are different (surname vs primary_name.surname_list[0].surname).

QueryGramplet.jpg


SQL-like:

select FIELDS from TABLE where PYTHON-BOOLEAN-EXPRESSION;

TABLES: FIELDS:

  1. people: given_name, surname, suffix, title, birth_date, death_date, gender, birth_place, death_place, change, marker
  2. families:
  3. sources:
  4. events:

Examples:

$ select * from people where surname.startswith("Smith")
QuerySmith.png


$ select given_name, surname from people;

$ select * from sources;
$ select * from events;
$ select * from families;