| 1 |
<!-- doc/src/sgml/advanced.sgml --> |
| 2 |
|
| 3 |
<chapter id="tutorial-advanced"> |
| 4 |
<title>Advanced Features</title> |
| 5 |
|
| 6 |
<sect1 id="tutorial-advanced-intro"> |
| 7 |
<title>Introduction</title> |
| 8 |
|
| 9 |
<para> |
| 10 |
In the previous chapter we have covered the basics of using |
| 11 |
<acronym>SQL</acronym> to store and access your data in |
| 12 |
<productname>PostgreSQL</productname>. We will now discuss some |
| 13 |
more advanced features of <acronym>SQL</acronym> that simplify |
| 14 |
management and prevent loss or corruption of your data. Finally, |
| 15 |
we will look at some <productname>PostgreSQL</productname> |
| 16 |
extensions. |
| 17 |
</para> |
| 18 |
|
| 19 |
<para> |
| 20 |
This chapter will on occasion refer to examples found in <xref |
| 21 |
linkend="tutorial-sql"/> to change or improve them, so it will be |
| 22 |
useful to have read that chapter. Some examples from |
| 23 |
this chapter can also be found in |
| 24 |
<filename>advanced.sql</filename> in the tutorial directory. This |
| 25 |
file also contains some sample data to load, which is not |
| 26 |
repeated here. (Refer to <xref linkend="tutorial-sql-intro"/> for |
| 27 |
how to use the file.) |
| 28 |
</para> |
| 29 |
</sect1> |
| 30 |
|
| 31 |
|
| 32 |
<sect1 id="tutorial-views"> |
| 33 |
<title>Views</title> |
| 34 |
|
| 35 |
<indexterm zone="tutorial-views"> |
| 36 |
<primary>view</primary> |
| 37 |
</indexterm> |
| 38 |
|
| 39 |
<para> |
| 40 |
Refer back to the queries in <xref linkend="tutorial-join"/>. |
| 41 |
Suppose the combined listing of weather records and city location |
| 42 |
is of particular interest to your application, but you do not want |
| 43 |
to type the query each time you need it. You can create a |
| 44 |
<firstterm>view</firstterm> over the query, which gives a name to |
| 45 |
the query that you can refer to like an ordinary table: |
| 46 |
|
| 47 |
<programlisting> |
| 48 |
CREATE VIEW myview AS |
| 49 |
SELECT name, temp_lo, temp_hi, prcp, date, location |
| 50 |
FROM weather, cities |
| 51 |
WHERE city = name; |
| 52 |
|
| 53 |
SELECT * FROM myview; |
| 54 |
</programlisting> |
| 55 |
</para> |
| 56 |
|
| 57 |
<para> |
| 58 |
Making liberal use of views is a key aspect of good SQL database |
| 59 |
design. Views allow you to encapsulate the details of the |
| 60 |
structure of your tables, which might change as your application |
| 61 |
evolves, behind consistent interfaces. |
| 62 |
</para> |
| 63 |
|
| 64 |
<para> |
| 65 |
Views can be used in almost any place a real table can be used. |
| 66 |
Building views upon other views is not uncommon. |
| 67 |
</para> |
| 68 |
</sect1> |
| 69 |
|
| 70 |
|
| 71 |
<sect1 id="tutorial-fk"> |
| 72 |
<title>Foreign Keys</title> |
| 73 |
|
| 74 |
<indexterm zone="tutorial-fk"> |
| 75 |
<primary>foreign key</primary> |
| 76 |
</indexterm> |
| 77 |
|
| 78 |
<indexterm zone="tutorial-fk"> |
| 79 |
<primary>referential integrity</primary> |
| 80 |
</indexterm> |
| 81 |
|
| 82 |
<para> |
| 83 |
Recall the <classname>weather</classname> and |
| 84 |
<classname>cities</classname> tables from <xref |
| 85 |
linkend="tutorial-sql"/>. Consider the following problem: You |
| 86 |
want to make sure that no one can insert rows in the |
| 87 |
<classname>weather</classname> table that do not have a matching |
| 88 |
entry in the <classname>cities</classname> table. This is called |
| 89 |
maintaining the <firstterm>referential integrity</firstterm> of |
| 90 |
your data. In simplistic database systems this would be |
| 91 |
implemented (if at all) by first looking at the |
| 92 |
<classname>cities</classname> table to check if a matching record |
| 93 |
exists, and then inserting or rejecting the new |
| 94 |
<classname>weather</classname> records. This approach has a |
| 95 |
number of problems and is very inconvenient, so |
| 96 |
<productname>PostgreSQL</productname> can do this for you. |
| 97 |
</para> |
| 98 |
|
| 99 |
<para> |
| 100 |
The new declaration of the tables would look like this: |
| 101 |
|
| 102 |
<programlisting> |
| 103 |
CREATE TABLE cities ( |
| 104 |
name varchar(80) primary key, |
| 105 |
location point |
| 106 |
); |
| 107 |
|
| 108 |
CREATE TABLE weather ( |
| 109 |
city varchar(80) references cities(name), |
| 110 |
temp_lo int, |
| 111 |
temp_hi int, |
| 112 |
prcp real, |
| 113 |
date date |
| 114 |
); |
| 115 |
</programlisting> |
| 116 |
|
| 117 |
Now try inserting an invalid record: |
| 118 |
|
| 119 |
<programlisting> |
| 120 |
INSERT INTO weather VALUES ('Berkeley', 45, 53, 0.0, '1994-11-28'); |
| 121 |
</programlisting> |
| 122 |
|
| 123 |
<screen> |
| 124 |
ERROR: insert or update on table "weather" violates foreign key constraint "weather_city_fkey" |
| 125 |
DETAIL: Key (city)=(Berkeley) is not present in table "cities". |
| 126 |
</screen> |
| 127 |
</para> |
| 128 |
|
| 129 |
<para> |
| 130 |
The behavior of foreign keys can be finely tuned to your |
| 131 |
application. We will not go beyond this simple example in this |
| 132 |
tutorial, but just refer you to <xref linkend="ddl"/> |
| 133 |
for more information. Making correct use of |
| 134 |
foreign keys will definitely improve the quality of your database |
| 135 |
applications, so you are strongly encouraged to learn about them. |
| 136 |
</para> |
| 137 |
</sect1> |
| 138 |
|
| 139 |
|
| 140 |
<sect1 id="tutorial-transactions"> |
| 141 |
<title>Transactions</title> |
| 142 |
|
| 143 |
<indexterm zone="tutorial-transactions"> |
| 144 |
<primary>transaction</primary> |
| 145 |
</indexterm> |
| 146 |
|
| 147 |
<para> |
| 148 |
<firstterm>Transactions</firstterm> are a fundamental concept of all database |
| 149 |
systems. The essential point of a transaction is that it bundles |
| 150 |
multiple steps into a single, all-or-nothing operation. The intermediate |
| 151 |
states between the steps are not visible to other concurrent transactions, |
| 152 |
and if some failure occurs that prevents the transaction from completing, |
| 153 |
then none of the steps affect the database at all. |
| 154 |
</para> |
| 155 |
|
| 156 |
<para> |
| 157 |
For example, consider a bank database that contains balances for various |
| 158 |
customer accounts, as well as total deposit balances for branches. |
| 159 |
Suppose that we want to record a payment of $100.00 from Alice's account |
| 160 |
to Bob's account. Simplifying outrageously, the SQL commands for this |
| 161 |
might look like: |
| 162 |
|
| 163 |
<programlisting> |
| 164 |
UPDATE accounts SET balance = balance - 100.00 |
| 165 |
WHERE name = 'Alice'; |
| 166 |
UPDATE branches SET balance = balance - 100.00 |
| 167 |
WHERE name = (SELECT branch_name FROM accounts WHERE name = 'Alice'); |
| 168 |
UPDATE accounts SET balance = balance + 100.00 |
| 169 |
WHERE name = 'Bob'; |
| 170 |
UPDATE branches SET balance = balance + 100.00 |
| 171 |
WHERE name = (SELECT branch_name FROM accounts WHERE name = 'Bob'); |
| 172 |
</programlisting> |
| 173 |
</para> |
| 174 |
|
| 175 |
<para> |
| 176 |
The details of these commands are not important here; the important |
| 177 |
point is that there are several separate updates involved to accomplish |
| 178 |
this rather simple operation. Our bank's officers will want to be |
| 179 |
assured that either all these updates happen, or none of them happen. |
| 180 |
It would certainly not do for a system failure to result in Bob |
| 181 |
receiving $100.00 that was not debited from Alice. Nor would Alice long |
| 182 |
remain a happy customer if she was debited without Bob being credited. |
| 183 |
We need a guarantee that if something goes wrong partway through the |
| 184 |
operation, none of the steps executed so far will take effect. Grouping |
| 185 |
the updates into a <firstterm>transaction</firstterm> gives us this guarantee. |
| 186 |
A transaction is said to be <firstterm>atomic</firstterm>: from the point of |
| 187 |
view of other transactions, it either happens completely or not at all. |
| 188 |
</para> |
| 189 |
|
| 190 |
<para> |
| 191 |
We also want a |
| 192 |
guarantee that once a transaction is completed and acknowledged by |
| 193 |
the database system, it has indeed been permanently recorded |
| 194 |
and won't be lost even if a crash ensues shortly thereafter. |
| 195 |
For example, if we are recording a cash withdrawal by Bob, |
| 196 |
we do not want any chance that the debit to his account will |
| 197 |
disappear in a crash just after he walks out the bank door. |
| 198 |
A transactional database guarantees that all the updates made by |
| 199 |
a transaction are logged in permanent storage (i.e., on disk) before |
| 200 |
the transaction is reported complete. |
| 201 |
</para> |
| 202 |
|
| 203 |
<para> |
| 204 |
Another important property of transactional databases is closely |
| 205 |
related to the notion of atomic updates: when multiple transactions |
| 206 |
are running concurrently, each one should not be able to see the |
| 207 |
incomplete changes made by others. For example, if one transaction |
| 208 |
is busy totalling all the branch balances, it would not do for it |
| 209 |
to include the debit from Alice's branch but not the credit to |
| 210 |
Bob's branch, nor vice versa. So transactions must be all-or-nothing |
| 211 |
not only in terms of their permanent effect on the database, but |
| 212 |
also in terms of their visibility as they happen. The updates made |
| 213 |
so far by an open transaction are invisible to other transactions |
| 214 |
until the transaction completes, whereupon all the updates become |
| 215 |
visible simultaneously. |
| 216 |
</para> |
| 217 |
|
| 218 |
<para> |
| 219 |
In <productname>PostgreSQL</productname>, a transaction is set up by surrounding |
| 220 |
the SQL commands of the transaction with |
| 221 |
<command>BEGIN</command> and <command>COMMIT</command> commands. So our banking |
| 222 |
transaction would actually look like: |
| 223 |
|
| 224 |
<programlisting> |
| 225 |
BEGIN; |
| 226 |
UPDATE accounts SET balance = balance - 100.00 |
| 227 |
WHERE name = 'Alice'; |
| 228 |
-- etc etc |
| 229 |
COMMIT; |
| 230 |
</programlisting> |
| 231 |
</para> |
| 232 |
|
| 233 |
<para> |
| 234 |
If, partway through the transaction, we decide we do not want to |
| 235 |
commit (perhaps we just noticed that Alice's balance went negative), |
| 236 |
we can issue the command <command>ROLLBACK</command> instead of |
| 237 |
<command>COMMIT</command>, and all our updates so far will be canceled. |
| 238 |
</para> |
| 239 |
|
| 240 |
<para> |
| 241 |
<productname>PostgreSQL</productname> actually treats every SQL statement as being |
| 242 |
executed within a transaction. If you do not issue a <command>BEGIN</command> |
| 243 |
command, |
| 244 |
then each individual statement has an implicit <command>BEGIN</command> and |
| 245 |
(if successful) <command>COMMIT</command> wrapped around it. A group of |
| 246 |
statements surrounded by <command>BEGIN</command> and <command>COMMIT</command> |
| 247 |
is sometimes called a <firstterm>transaction block</firstterm>. |
| 248 |
</para> |
| 249 |
|
| 250 |
<note> |
| 251 |
<para> |
| 252 |
Some client libraries issue <command>BEGIN</command> and <command>COMMIT</command> |
| 253 |
commands automatically, so that you might get the effect of transaction |
| 254 |
blocks without asking. Check the documentation for the interface |
| 255 |
you are using. |
| 256 |
</para> |
| 257 |
</note> |
| 258 |
|
| 259 |
<para> |
| 260 |
It's possible to control the statements in a transaction in a more |
| 261 |
granular fashion through the use of <firstterm>savepoints</firstterm>. Savepoints |
| 262 |
allow you to selectively discard parts of the transaction, while |
| 263 |
committing the rest. After defining a savepoint with |
| 264 |
<command>SAVEPOINT</command>, you can if needed roll back to the savepoint |
| 265 |
with <command>ROLLBACK TO</command>. All the transaction's database changes |
| 266 |
between defining the savepoint and rolling back to it are discarded, but |
| 267 |
changes earlier than the savepoint are kept. |
| 268 |
</para> |
| 269 |
|
| 270 |
<para> |
| 271 |
After rolling back to a savepoint, it continues to be defined, so you can |
| 272 |
roll back to it several times. Conversely, if you are sure you won't need |
| 273 |
to roll back to a particular savepoint again, it can be released, so the |
| 274 |
system can free some resources. Keep in mind that either releasing or |
| 275 |
rolling back to a savepoint |
| 276 |
will automatically release all savepoints that were defined after it. |
| 277 |
</para> |
| 278 |
|
| 279 |
<para> |
| 280 |
All this is happening within the transaction block, so none of it |
| 281 |
is visible to other database sessions. When and if you commit the |
| 282 |
transaction block, the committed actions become visible as a unit |
| 283 |
to other sessions, while the rolled-back actions never become visible |
| 284 |
at all. |
| 285 |
</para> |
| 286 |
|
| 287 |
<para> |
| 288 |
Remembering the bank database, suppose we debit $100.00 from Alice's |
| 289 |
account, and credit Bob's account, only to find later that we should |
| 290 |
have credited Wally's account. We could do it using savepoints like |
| 291 |
this: |
| 292 |
|
| 293 |
<programlisting> |
| 294 |
BEGIN; |
| 295 |
UPDATE accounts SET balance = balance - 100.00 |
| 296 |
WHERE name = 'Alice'; |
| 297 |
SAVEPOINT my_savepoint; |
| 298 |
UPDATE accounts SET balance = balance + 100.00 |
| 299 |
WHERE name = 'Bob'; |
| 300 |
-- oops ... forget that and use Wally's account |
| 301 |
ROLLBACK TO my_savepoint; |
| 302 |
UPDATE accounts SET balance = balance + 100.00 |
| 303 |
WHERE name = 'Wally'; |
| 304 |
COMMIT; |
| 305 |
</programlisting> |
| 306 |
</para> |
| 307 |
|
| 308 |
<para> |
| 309 |
This example is, of course, oversimplified, but there's a lot of control |
| 310 |
possible in a transaction block through the use of savepoints. |
| 311 |
Moreover, <command>ROLLBACK TO</command> is the only way to regain control of a |
| 312 |
transaction block that was put in aborted state by the |
| 313 |
system due to an error, short of rolling it back completely and starting |
| 314 |
again. |
| 315 |
</para> |
| 316 |
|
| 317 |
</sect1> |
| 318 |
|
| 319 |
|
| 320 |
<sect1 id="tutorial-window"> |
| 321 |
<title>Window Functions</title> |
| 322 |
|
| 323 |
<indexterm zone="tutorial-window"> |
| 324 |
<primary>window function</primary> |
| 325 |
</indexterm> |
| 326 |
|
| 327 |
<para> |
| 328 |
A <firstterm>window function</firstterm> performs a calculation across a set of |
| 329 |
table rows that are somehow related to the current row. This is comparable |
| 330 |
to the type of calculation that can be done with an aggregate function. |
| 331 |
However, window functions do not cause rows to become grouped into a single |
| 332 |
output row like non-window aggregate calls would. Instead, the |
| 333 |
rows retain their separate identities. Behind the scenes, the window |
| 334 |
function is able to access more than just the current row of the query |
| 335 |
result. |
| 336 |
</para> |
| 337 |
|
| 338 |
<para> |
| 339 |
Here is an example that shows how to compare each employee's salary |
| 340 |
with the average salary in his or her department: |
| 341 |
|
| 342 |
<programlisting> |
| 343 |
SELECT depname, empno, salary, avg(salary) OVER (PARTITION BY depname) FROM empsalary; |
| 344 |
</programlisting> |
| 345 |
|
| 346 |
<screen> |
| 347 |
depname | empno | salary | avg |
| 348 |
-----------+-------+--------+----------------------- |
| 349 |
develop | 11 | 5200 | 5020.0000000000000000 |
| 350 |
develop | 7 | 4200 | 5020.0000000000000000 |
| 351 |
develop | 9 | 4500 | 5020.0000000000000000 |
| 352 |
develop | 8 | 6000 | 5020.0000000000000000 |
| 353 |
develop | 10 | 5200 | 5020.0000000000000000 |
| 354 |
personnel | 5 | 3500 | 3700.0000000000000000 |
| 355 |
personnel | 2 | 3900 | 3700.0000000000000000 |
| 356 |
sales | 3 | 4800 | 4866.6666666666666667 |
| 357 |
sales | 1 | 5000 | 4866.6666666666666667 |
| 358 |
sales | 4 | 4800 | 4866.6666666666666667 |
| 359 |
(10 rows) |
| 360 |
</screen> |
| 361 |
|
| 362 |
The first three output columns come directly from the table |
| 363 |
<structname>empsalary</structname>, and there is one output row for each row in the |
| 364 |
table. The fourth column represents an average taken across all the table |
| 365 |
rows that have the same <structfield>depname</structfield> value as the current row. |
| 366 |
(This actually is the same function as the non-window <function>avg</function> |
| 367 |
aggregate, but the <literal>OVER</literal> clause causes it to be |
| 368 |
treated as a window function and computed across the window frame.) |
| 369 |
</para> |
| 370 |
|
| 371 |
<para> |
| 372 |
A window function call always contains an <literal>OVER</literal> clause |
| 373 |
directly following the window function's name and argument(s). This is what |
| 374 |
syntactically distinguishes it from a normal function or non-window |
| 375 |
aggregate. The <literal>OVER</literal> clause determines exactly how the |
| 376 |
rows of the query are split up for processing by the window function. |
| 377 |
The <literal>PARTITION BY</literal> clause within <literal>OVER</literal> |
| 378 |
divides the rows into groups, or partitions, that share the same |
| 379 |
values of the <literal>PARTITION BY</literal> expression(s). For each row, |
| 380 |
the window function is computed across the rows that fall into the |
| 381 |
same partition as the current row. |
| 382 |
</para> |
| 383 |
|
| 384 |
<para> |
| 385 |
You can also control the order in which rows are processed by |
| 386 |
window functions using <literal>ORDER BY</literal> within <literal>OVER</literal>. |
| 387 |
(The window <literal>ORDER BY</literal> does not even have to match the |
| 388 |
order in which the rows are output.) Here is an example: |
| 389 |
|
| 390 |
<programlisting> |
| 391 |
SELECT depname, empno, salary, |
| 392 |
row_number() OVER (PARTITION BY depname ORDER BY salary DESC) |
| 393 |
FROM empsalary; |
| 394 |
</programlisting> |
| 395 |
|
| 396 |
<screen> |
| 397 |
depname | empno | salary | row_number |
| 398 |
-----------+-------+--------+------------ |
| 399 |
develop | 8 | 6000 | 1 |
| 400 |
develop | 10 | 5200 | 2 |
| 401 |
develop | 11 | 5200 | 3 |
| 402 |
develop | 9 | 4500 | 4 |
| 403 |
develop | 7 | 4200 | 5 |
| 404 |
personnel | 2 | 3900 | 1 |
| 405 |
personnel | 5 | 3500 | 2 |
| 406 |
sales | 1 | 5000 | 1 |
| 407 |
sales | 4 | 4800 | 2 |
| 408 |
sales | 3 | 4800 | 3 |
| 409 |
(10 rows) |
| 410 |
</screen> |
| 411 |
|
| 412 |
As shown here, the <function>row_number</function> window function |
| 413 |
assigns sequential numbers to the rows within each partition, |
| 414 |
in the order defined by the <literal>ORDER BY</literal> clause |
| 415 |
(with tied rows numbered in an unspecified order). |
| 416 |
<function>row_number</function> needs no explicit parameter, |
| 417 |
because its behavior |
| 418 |
is entirely determined by the <literal>OVER</literal> clause. |
| 419 |
</para> |
| 420 |
|
| 421 |
<para> |
| 422 |
The rows considered by a window function are those of the <quote>virtual |
| 423 |
table</quote> produced by the query's <literal>FROM</literal> clause as filtered by its |
| 424 |
<literal>WHERE</literal>, <literal>GROUP BY</literal>, and <literal>HAVING</literal> clauses |
| 425 |
if any. For example, a row removed because it does not meet the |
| 426 |
<literal>WHERE</literal> condition is not seen by any window function. |
| 427 |
A query can contain multiple window functions that slice up the data |
| 428 |
in different ways using different <literal>OVER</literal> clauses, but |
| 429 |
they all act on the same collection of rows defined by this virtual table. |
| 430 |
</para> |
| 431 |
|
| 432 |
<para> |
| 433 |
We already saw that <literal>ORDER BY</literal> can be omitted if the ordering |
| 434 |
of rows is not important. It is also possible to omit <literal>PARTITION |
| 435 |
BY</literal>, in which case there is a single partition containing all rows. |
| 436 |
</para> |
| 437 |
|
| 438 |
<para> |
| 439 |
There is another important concept associated with window functions: |
| 440 |
for each row, there is a set of rows within its partition called its |
| 441 |
<firstterm>window frame</firstterm>. Some window functions act only |
| 442 |
on the rows of the window frame, rather than of the whole partition. |
| 443 |
By default, if <literal>ORDER BY</literal> is supplied then the frame consists of |
| 444 |
all rows from the start of the partition up through the current row, plus |
| 445 |
any following rows that are equal to the current row according to the |
| 446 |
<literal>ORDER BY</literal> clause. When <literal>ORDER BY</literal> is omitted the |
| 447 |
default frame consists of all rows in the partition. |
| 448 |
<footnote> |
| 449 |
<para> |
| 450 |
There are options to define the window frame in other ways, but |
| 451 |
this tutorial does not cover them. See |
| 452 |
<xref linkend="syntax-window-functions"/> for details. |
| 453 |
</para> |
| 454 |
</footnote> |
| 455 |
Here is an example using <function>sum</function>: |
| 456 |
</para> |
| 457 |
|
| 458 |
<programlisting> |
| 459 |
SELECT salary, sum(salary) OVER () FROM empsalary; |
| 460 |
</programlisting> |
| 461 |
|
| 462 |
<screen> |
| 463 |
salary | sum |
| 464 |
--------+------- |
| 465 |
5200 | 47100 |
| 466 |
5000 | 47100 |
| 467 |
3500 | 47100 |
| 468 |
4800 | 47100 |
| 469 |
3900 | 47100 |
| 470 |
4200 | 47100 |
| 471 |
4500 | 47100 |
| 472 |
4800 | 47100 |
| 473 |
6000 | 47100 |
| 474 |
5200 | 47100 |
| 475 |
(10 rows) |
| 476 |
</screen> |
| 477 |
|
| 478 |
<para> |
| 479 |
Above, since there is no <literal>ORDER BY</literal> in the <literal>OVER</literal> |
| 480 |
clause, the window frame is the same as the partition, which for lack of |
| 481 |
<literal>PARTITION BY</literal> is the whole table; in other words each sum is |
| 482 |
taken over the whole table and so we get the same result for each output |
| 483 |
row. But if we add an <literal>ORDER BY</literal> clause, we get very different |
| 484 |
results: |
| 485 |
</para> |
| 486 |
|
| 487 |
<programlisting> |
| 488 |
SELECT salary, sum(salary) OVER (ORDER BY salary) FROM empsalary; |
| 489 |
</programlisting> |
| 490 |
|
| 491 |
<screen> |
| 492 |
salary | sum |
| 493 |
--------+------- |
| 494 |
3500 | 3500 |
| 495 |
3900 | 7400 |
| 496 |
4200 | 11600 |
| 497 |
4500 | 16100 |
| 498 |
4800 | 25700 |
| 499 |
4800 | 25700 |
| 500 |
5000 | 30700 |
| 501 |
5200 | 41100 |
| 502 |
5200 | 41100 |
| 503 |
6000 | 47100 |
| 504 |
(10 rows) |
| 505 |
</screen> |
| 506 |
|
| 507 |
<para> |
| 508 |
Here the sum is taken from the first (lowest) salary up through the |
| 509 |
current one, including any duplicates of the current one (notice the |
| 510 |
results for the duplicated salaries). |
| 511 |
</para> |
| 512 |
|
| 513 |
<para> |
| 514 |
Window functions are permitted only in the <literal>SELECT</literal> list |
| 515 |
and the <literal>ORDER BY</literal> clause of the query. They are forbidden |
| 516 |
elsewhere, such as in <literal>GROUP BY</literal>, <literal>HAVING</literal> |
| 517 |
and <literal>WHERE</literal> clauses. This is because they logically |
| 518 |
execute after the processing of those clauses. Also, window functions |
| 519 |
execute after non-window aggregate functions. This means it is valid to |
| 520 |
include an aggregate function call in the arguments of a window function, |
| 521 |
but not vice versa. |
| 522 |
</para> |
| 523 |
|
| 524 |
<para> |
| 525 |
If there is a need to filter or group rows after the window calculations |
| 526 |
are performed, you can use a sub-select. For example: |
| 527 |
|
| 528 |
<programlisting> |
| 529 |
SELECT depname, empno, salary, enroll_date |
| 530 |
FROM |
| 531 |
(SELECT depname, empno, salary, enroll_date, |
| 532 |
row_number() OVER (PARTITION BY depname ORDER BY salary DESC, empno) AS pos |
| 533 |
FROM empsalary |
| 534 |
) AS ss |
| 535 |
WHERE pos < 3; |
| 536 |
</programlisting> |
| 537 |
|
| 538 |
The above query only shows the rows from the inner query having |
| 539 |
<literal>row_number</literal> less than 3 (that is, the first |
| 540 |
two rows for each department). |
| 541 |
</para> |
| 542 |
|
| 543 |
<para> |
| 544 |
When a query involves multiple window functions, it is possible to write |
| 545 |
out each one with a separate <literal>OVER</literal> clause, but this is |
| 546 |
duplicative and error-prone if the same windowing behavior is wanted |
| 547 |
for several functions. Instead, each windowing behavior can be named |
| 548 |
in a <literal>WINDOW</literal> clause and then referenced in <literal>OVER</literal>. |
| 549 |
For example: |
| 550 |
|
| 551 |
<programlisting> |
| 552 |
SELECT sum(salary) OVER w, avg(salary) OVER w |
| 553 |
FROM empsalary |
| 554 |
WINDOW w AS (PARTITION BY depname ORDER BY salary DESC); |
| 555 |
</programlisting> |
| 556 |
</para> |
| 557 |
|
| 558 |
<para> |
| 559 |
More details about window functions can be found in |
| 560 |
<xref linkend="syntax-window-functions"/>, |
| 561 |
<xref linkend="functions-window"/>, |
| 562 |
<xref linkend="queries-window"/>, and the |
| 563 |
<xref linkend="sql-select"/> reference page. |
| 564 |
</para> |
| 565 |
</sect1> |
| 566 |
|
| 567 |
|
| 568 |
<sect1 id="tutorial-inheritance"> |
| 569 |
<title>Inheritance</title> |
| 570 |
|
| 571 |
<indexterm zone="tutorial-inheritance"> |
| 572 |
<primary>inheritance</primary> |
| 573 |
</indexterm> |
| 574 |
|
| 575 |
<para> |
| 576 |
Inheritance is a concept from object-oriented databases. It opens |
| 577 |
up interesting new possibilities of database design. |
| 578 |
</para> |
| 579 |
|
| 580 |
<para> |
| 581 |
Let's create two tables: A table <classname>cities</classname> |
| 582 |
and a table <classname>capitals</classname>. Naturally, capitals |
| 583 |
are also cities, so you want some way to show the capitals |
| 584 |
implicitly when you list all cities. If you're really clever you |
| 585 |
might invent some scheme like this: |
| 586 |
|
| 587 |
<programlisting> |
| 588 |
CREATE TABLE capitals ( |
| 589 |
name text, |
| 590 |
population real, |
| 591 |
elevation int, -- (in ft) |
| 592 |
state char(2) |
| 593 |
); |
| 594 |
|
| 595 |
CREATE TABLE non_capitals ( |
| 596 |
name text, |
| 597 |
population real, |
| 598 |
elevation int -- (in ft) |
| 599 |
); |
| 600 |
|
| 601 |
CREATE VIEW cities AS |
| 602 |
SELECT name, population, elevation FROM capitals |
| 603 |
UNION |
| 604 |
SELECT name, population, elevation FROM non_capitals; |
| 605 |
</programlisting> |
| 606 |
|
| 607 |
This works OK as far as querying goes, but it gets ugly when you |
| 608 |
need to update several rows, for one thing. |
| 609 |
</para> |
| 610 |
|
| 611 |
<para> |
| 612 |
A better solution is this: |
| 613 |
|
| 614 |
<programlisting> |
| 615 |
CREATE TABLE cities ( |
| 616 |
name text, |
| 617 |
population real, |
| 618 |
elevation int -- (in ft) |
| 619 |
); |
| 620 |
|
| 621 |
CREATE TABLE capitals ( |
| 622 |
state char(2) UNIQUE NOT NULL |
| 623 |
) INHERITS (cities); |
| 624 |
</programlisting> |
| 625 |
</para> |
| 626 |
|
| 627 |
<para> |
| 628 |
In this case, a row of <classname>capitals</classname> |
| 629 |
<firstterm>inherits</firstterm> all columns (<structfield>name</structfield>, |
| 630 |
<structfield>population</structfield>, and <structfield>elevation</structfield>) from its |
| 631 |
<firstterm>parent</firstterm>, <classname>cities</classname>. The |
| 632 |
type of the column <structfield>name</structfield> is |
| 633 |
<type>text</type>, a native <productname>PostgreSQL</productname> |
| 634 |
type for variable length character strings. The |
| 635 |
<classname>capitals</classname> table has |
| 636 |
an additional column, <structfield>state</structfield>, which shows its |
| 637 |
state abbreviation. In |
| 638 |
<productname>PostgreSQL</productname>, a table can inherit from |
| 639 |
zero or more other tables. |
| 640 |
</para> |
| 641 |
|
| 642 |
<para> |
| 643 |
For example, the following query finds the names of all cities, |
| 644 |
including state capitals, that are located at an elevation |
| 645 |
over 500 feet: |
| 646 |
|
| 647 |
<programlisting> |
| 648 |
SELECT name, elevation |
| 649 |
FROM cities |
| 650 |
WHERE elevation > 500; |
| 651 |
</programlisting> |
| 652 |
|
| 653 |
which returns: |
| 654 |
|
| 655 |
<screen> |
| 656 |
name | elevation |
| 657 |
-----------+----------- |
| 658 |
Las Vegas | 2174 |
| 659 |
Mariposa | 1953 |
| 660 |
Madison | 845 |
| 661 |
(3 rows) |
| 662 |
</screen> |
| 663 |
</para> |
| 664 |
|
| 665 |
<para> |
| 666 |
On the other hand, the following query finds |
| 667 |
all the cities that are not state capitals and |
| 668 |
are situated at an elevation over 500 feet: |
| 669 |
|
| 670 |
<programlisting> |
| 671 |
SELECT name, elevation |
| 672 |
FROM ONLY cities |
| 673 |
WHERE elevation > 500; |
| 674 |
</programlisting> |
| 675 |
|
| 676 |
<screen> |
| 677 |
name | elevation |
| 678 |
-----------+----------- |
| 679 |
Las Vegas | 2174 |
| 680 |
Mariposa | 1953 |
| 681 |
(2 rows) |
| 682 |
</screen> |
| 683 |
</para> |
| 684 |
|
| 685 |
<para> |
| 686 |
Here the <literal>ONLY</literal> before <literal>cities</literal> |
| 687 |
indicates that the query should be run over only the |
| 688 |
<classname>cities</classname> table, and not tables below |
| 689 |
<classname>cities</classname> in the inheritance hierarchy. Many |
| 690 |
of the commands that we have already discussed — |
| 691 |
<command>SELECT</command>, <command>UPDATE</command>, and |
| 692 |
<command>DELETE</command> — support this <literal>ONLY</literal> |
| 693 |
notation. |
| 694 |
</para> |
| 695 |
|
| 696 |
<note> |
| 697 |
<para> |
| 698 |
Although inheritance is frequently useful, it has not been integrated |
| 699 |
with unique constraints or foreign keys, which limits its usefulness. |
| 700 |
See <xref linkend="ddl-inherit"/> for more detail. |
| 701 |
</para> |
| 702 |
</note> |
| 703 |
</sect1> |
| 704 |
|
| 705 |
|
| 706 |
<sect1 id="tutorial-conclusion"> |
| 707 |
<title>Conclusion</title> |
| 708 |
|
| 709 |
<para> |
| 710 |
<productname>PostgreSQL</productname> has many features not |
| 711 |
touched upon in this tutorial introduction, which has been |
| 712 |
oriented toward newer users of <acronym>SQL</acronym>. These |
| 713 |
features are discussed in more detail in the remainder of this |
| 714 |
book. |
| 715 |
</para> |
| 716 |
|
| 717 |
<para> |
| 718 |
If you feel you need more introductory material, please visit the PostgreSQL |
| 719 |
<ulink url="https://www.postgresql.org">web site</ulink> |
| 720 |
for links to more resources. |
| 721 |
</para> |
| 722 |
</sect1> |
| 723 |
</chapter> |