Combing "previous row" of same table and JOIN from different table in Sqlite

I have the following table

CREATE TABLE "shots" (
    "player"    INTEGER,
    "tournament"    TEXT,
    "year"  INTEGER,
    "course"    INTEGER,
    "round" INTEGER,
    "hole"  INTEGER,
    "shot"  INTEGER,
    "text"  TEXT,
    "distance"  REAL,
    "x" TEXT,
    "y" TEXT,
    "z" TEXT
);

With a sample of the data:

28237   470 2015    717 1   1   1   Shot 1 302 yds to left fairway, 257 yds to hole 10874   11451.596   10623.774   78.251
28237   470 2015    717 1   1   2   Shot 2 234 yds to right fairway, 71 ft to hole  8437    12150.454   10700.381   86.035
28237   470 2015    717 1   1   3   Shot 3 70 ft to green, 4 ft to hole 838 12215.728   10725.134   88.408
28237   470 2015    717 1   1   4   Shot 4 in the hole  46  12215.1 10729.1 88.371
28237   470 2015    717 1   2   1   Shot 1 199 yds to green, 29 ft to hole  7162    12776.03    10398.086   91.017
28237   470 2015    717 1   2   2   Shot 2 putt 26 ft 7 in., 2 ft 4 in. to hole 319 12749.444   10398.854   90.998
28237   470 2015    717 1   2   3   Shot 3 in the hole  28  12747.3 10397.6 91.027
28237   470 2015    717 1   3   1   Shot 1 296 yds to left intermediate, 204 yds to hole    10651   12596.857   9448.27 94.296
28237   470 2015    717 1   3   2   Shot 2 208 yds to green, 15 ft to hole  7478    12571.0 8825.648    94.673
28237   470 2015    717 1   3   3   Shot 3 putt 17 ft 6 in., 2 ft 5 in. to hole 210 12561.831   8840.539    94.362

I want to get for each shot the previous location (x, y, z). I wrote the below query.

SELECT cur.player, cur.tournament, cur.year, cur.course, cur.round, cur.hole, cur.shot, cur.x, cur.y, cur.z, prev.x, prev.y, prev.z
FROM shots cur
INNER JOIN shots prev 
ON (cur.player, cur.tournament, cur.year, cur.course, cur.round, cur.hole, cur.shot) =
   (prev.player, prev.tournament, prev.year, prev.course, prev.round, prev.hole, prev.shot - 1)

This query takes forever basically. How can I rewrite it to make it faster? In addition, I need to make an adjustment for the first shot on a hole (shot = 1). This shot is made from tee_x, tee_y and tee_z. These values are available in table holes

CREATE TABLE "holes" (
        "tournament"    TEXT,
        "year"  INTEGER,
        "course"    INTEGER,
        "round" INTEGER,
        "hole"  INTEGER,
        "tee_x" TEXT,
        "tee_y" TEXT,
        "tee_z" TEXT
    );

With data:

   470  2015    717 1   1   11450   10625   78.25
   470  2015    717 1   2   12750   10400   91
   470  2015    717 1   3   2565    8840.5  95

Thanks

Solution

First, you need a composite index to speed up the operation:

CREATE INDEX idx_shots ON shots (player, tournament, year, course, round, hole, shot);

With that index, your query should run faster:

SELECT cur.player, cur.tournament, cur.year, cur.course, cur.round, cur.hole, cur.shot, cur.x, cur.y, cur.z, 
       prev.x AS prev_x, prev.y AS prev_y, prev.z AS prev_z
FROM shots cur LEFT JOIN shots prev 
ON (cur.player, cur.tournament, cur.year, cur.course, cur.round, cur.hole, cur.shot) =
   (prev.player, prev.tournament, prev.year, prev.course, prev.round, prev.hole, prev.shot + 1);

The changes I made:

the join should be a LEFT join so that all rows are included and not only the ones that have a previous row
-1 should be +1 because the previous row's shot is 1 less than the current row's shot
added aliases for the previous row's x, y and z

But, if your version of SQLite is 3.25.0+ it would be better to use window function LAG() instead of a self join:

SELECT *,
       LAG(x) OVER w AS prev_x,
       LAG(y) OVER w AS prev_y,
       LAG(z) OVER w AS prev_z
FROM shots
WINDOW w AS (PARTITION BY player, tournament, year, course, round, hole ORDER BY shot);

See the demo (I include the query plan for both queries where you can see the use of the composite index).