existing.sql

create table male
  ( male_id  int           not null
  , name     varchar(100)  not null
  , primary key (male_id) );

create table female
  ( female_id  int           not null
  , name       varchar(100)  not null
  , primary key (female_id) );

create view person as
  select 'm' as gender, male_id as person_id, name from male
  union
  select 'f' as gender, female_id as person_id, name from female;

-- let's say we're given (@gender, @person_id, @new_name),
-- and we want to update the corresponding person's name
if @gender = 'm'
  update male set name = @new_name where male_id = @person_id
else if @gender = 'f'
  update female set name = @new_name where female_id = @person_id
else
  -- what do?? we've been given crap data, this should be impossible
  -- types are supposed to prevent this kind of error!

proposed.sql

-- every table has an autogenerated id field
-- for primitive tables, it's an abstract type (internally an autoincremental int)
-- for derived tables, it's an ML-like sum (variant) or product (tuple)

-- in my system, there are no null fields whatsoever,
-- so `not null` annotations are unnecessary

create table male
  primitive
  fields ( minus_factor numeric )

create table female
  primitive
  fields ( plus_factor numeric )

create table person
  derived = Male of male
          | Female of female
  fields ( name   varchar(100)
         , health numeric )

-- let's say we are given (@person, @new_name),
-- and we want to update the corresponding person's name
set @person.name := @new_name

-- let's say we are given (@person)
-- if it's a male, we want to decrease their health by 20%
-- if it's a female, we want to decrease their health by 10%
set @person.health := @person.health *
  (case @person of
      Male _ => 0.8
    | Female _ => 0.9)

-- let's say we want to decrease the health of all males by their minus_factor
-- and increase the health of all females by their plus_factor
forall [@person : person]
  set @person.health := @person.health *
    (case @person of
        Male @male => (1 - @male.minus_factor)
      | Female @female => (1 + @female.plus_factor))

-- alternatively we could do this:
forall [@male : male]
  let @person = Male @male in
  set @person.health := @person.health * (1 - @male.minus_factor)
forall [@female : female]
  let @person = Female @female in
  set @person.health := @person.health * (1 + @female.plus_factor)

Name table solution @ https://gist.github.com/spion/d60382ff2787967cb104