Select where number range does not overlap

问题

I have two tables that contain records about road construction activites:

• table_a is the master list.
• table_b is a legacy list.

For each road, in each year, I want to select the records from table_b that do not already exist in table_a.

Also, the records should not overlap spatially along the road. More specifically, the from_m and to_m of the records in table_b should not overlap the from_m and to_m in table_a.

How can I do this? I do not have Oracle Spatial.

---

The data in Excel (for easy viewing):

Here is what the data looks like in Excel:

The records in green should be selected by the query; the records in red should not.

---

The DDL:

Table A:

  create table table_a 
   (
    id number(4,0), 
    road_id number(4,0), 
    year number(4,0), 
    from_m number(4,0), 
    to_m number(4,0)
   );

insert into table_a (id,road_id,year,from_m,to_m) values (1,1,2000,0,100);
insert into table_a (id,road_id,year,from_m,to_m) values (2,1,2005,0,25);
insert into table_a (id,road_id,year,from_m,to_m) values (3,1,2005,50,75);
insert into table_a (id,road_id,year,from_m,to_m) values (4,1,2005,75,100);
insert into table_a (id,road_id,year,from_m,to_m) values (5,1,2010,10,50);
insert into table_a (id,road_id,year,from_m,to_m) values (6,1,2010,50,90);
insert into table_a (id,road_id,year,from_m,to_m) values (7,1,2015,40,100);
insert into table_a (id,road_id,year,from_m,to_m) values (8,2,2020,0,40);
insert into table_a (id,road_id,year,from_m,to_m) values (9,2,2020,0,40);
insert into table_a (id,road_id,year,from_m,to_m) values (10,3,2025,90,150);
commit;

select * from table_a;

        ID    ROAD_ID       YEAR     FROM_M       TO_M
---------- ---------- ---------- ---------- ----------
         1          1       2000          0        100
         2          1       2005          0         25
         3          1       2005         50         75
         4          1       2005         75        100
         5          1       2010         10         50
         6          1       2010         50         90
         7          1       2015         40        100
         8          2       2020          0         40
         9          2       2020          0         40
        10          3       2025         90        150

---

Table B:

  create table table_b 
   (
   id number(4,0), 
    road_id number(4,0), 
    year number(4,0), 
    from_m number(4,0), 
    to_m number(4,0)
   );

insert into table_b (id,road_id,year,from_m,to_m) values (1,1,1995,0,100);
insert into table_b (id,road_id,year,from_m,to_m) values (2,1,2001,0,50);
insert into table_b (id,road_id,year,from_m,to_m) values (3,1,2005,20,80);
insert into table_b (id,road_id,year,from_m,to_m) values (4,1,2005,0,100);
insert into table_b (id,road_id,year,from_m,to_m) values (5,1,2010,0,10);
insert into table_b (id,road_id,year,from_m,to_m) values (6,1,2010,90,100);
insert into table_b (id,road_id,year,from_m,to_m) values (7,1,2010,5,85);
insert into table_b (id,road_id,year,from_m,to_m) values (8,1,2015,40,100);
insert into table_b (id,road_id,year,from_m,to_m) values (9,1,2015,0,40);
insert into table_b (id,road_id,year,from_m,to_m) values (10,2,2020,0,41);
insert into table_b (id,road_id,year,from_m,to_m) values (11,3,2025,155,200);
insert into table_b (id,road_id,year,from_m,to_m) values (12,3,2025,199,300);
insert into table_b (id,road_id,year,from_m,to_m) values (13,4,2024,5,355);
commit;

select * from table_b;

        ID    ROAD_ID       YEAR     FROM_M       TO_M
---------- ---------- ---------- ---------- ----------
         1          1       1995          0        100
         2          1       2001          0         50
         3          1       2005         20         80
         4          1       2005          0        100
         5          1       2010          0         10
         6          1       2010         90        100
         7          1       2010          5         85
         8          1       2015         40        100
         9          1       2015          0         40
        10          2       2020          0         41
        11          3       2025        155        200
        12          3       2025        199        300
        13          4       2024          5        355

回答1:

A NOT EXISTS sub-select can help here

SELECT *
FROM table_b b
WHERE
    NOT EXISTS (SELECT *
                FROM table_a a
                WHERE
                    a.road_id = b.road_id AND
                    a.year = b.year AND
                    a.to_m > b.from_m AND
                    a.from_m < b.to_m)

Let's look at overlapping ranges (f=from, t=to)

a   -------------------f=======================t-----------------

b1a -----f=============t-----------------------------------------
b1b --f=============t--------------------------------------------

b2a -------------------------------------------f======t----------
b2b -----------------------------------------------f======t------

b3  ---------------f=========t-----------------------------------
b4  ------------------------f===========t------------------------
b5  ---------------------------------------f===========t---------

The ranges b3, b4 and b5 overlap. for all of them the following is true

a.to > b.from && a.from < b.to

For b1a, b1b and b2a, b2b that don't overlap this condition is false. For b1a a.from == b.to, for b1b a.from > b.to therefore the condition a.from < b.to is false.

For b2a a.to == b.from, for b2b a.to < b.from therefore the condition a.to > b.from is false.

The trick is to compare the from of one range with the to of the other one and vice-versa.

See: http://sqlfiddle.com/#!4/85883/3/0

回答2:

this will work :

SELECT * FROM table_b b
WHERE
    EXISTS (SELECT *
                FROM table_a a
                WHERE
                 a.year=b.year and a.from_m<>b.from_m and  a.to_m <>b.to_m );

来源：https://stackoverflow.com/questions/51729896/select-where-number-range-does-not-overlap