问题
I've an Oracle database with two schemas. One is old and another is new. I would like to update the old schema with the new columns of the new schema.
I get the tables which have changes with the following query.
select distinct table_name
from
(
select table_name,column_name
from all_tab_cols
where owner = 'SCHEMA_1'
minus
select table_name,column_name
from all_tab_cols
where owner = 'SCHEMA_2'
)
With this query I get the tables. How can I update the old schema tables with the new columns? I don't need the data, just the columns.
回答1:
A schema comparison tool is a good idea. The database schema is far more complicated than most people give credit, and every difference between two database schemas has the potential to cause bugs.
If you're still keen to do it yourself, the best approach I've found is to extract the schema definitions to text, then run a text compare. As long as everything is sorted alphabetically, you can then use Compare Documents feature in Microsoft Word (or FC.EXE, DIFF or equivalent), to highlight the differences.
The following SQLPlus script outputs the schema definition alphabetically, to allow comparison. There are two sections. The first section lists each column, in the format:
table_name.column_name: data_type = data_default <nullable>
The second section lists indexes and constraints, as follows:
PK constraint_name on table_name (pk_column_list)
FK constraint_name on table_name (fk_column_list)
CHECK constraint_name on table_name (constraint_definition)
The script serves as a useful references for extracting some of the Oracle schema details. This can be good knowledge to have when you're out at client sites and you don't have your usual tools available, or when security policies prevent you from accessing a client site database directly from your own PC.
set serveroutput on;
set serveroutput on size 1000000;
declare
rowcnt pls_integer := 0;
cursor c_column is
select table_name, column_name, data_type,
data_precision, data_length, data_scale,
data_default, nullable,
decode(data_scale, null, null, ',') scale_comma,
decode(default_length, null, null, '= ') default_equals
from all_tab_columns where owner = 'BCC'
order by table_name, column_name;
cursor c_constraint is
select c.table_name, c.constraint_name,
decode(c.constraint_type,
'P','PK',
'R','FK',
'C','CHECK',
c.constraint_type) constraint_type,
c.search_condition,
cc.column_1||cc.comma_2||cc.column_2||cc.comma_3||cc.column_3||cc.comma_4||cc.column_4||
cc.comma_5||cc.column_5||cc.comma_6||cc.column_6||cc.comma_7||cc.column_7 r_columns
from all_constraints c,
( select owner, table_name, constraint_name, nvl(max(position),0) max_position,
max( decode( position, 1, column_name, null ) ) column_1,
max( decode( position, 2, decode(column_name, null, null, ',' ), null ) ) comma_2,
max( decode( position, 2, column_name, null ) ) column_2,
max( decode( position, 3, decode(column_name, null, null, ',' ), null ) ) comma_3,
max( decode( position, 3, column_name, null ) ) column_3,
max( decode( position, 4, decode(column_name, null, null, ',' ), null ) ) comma_4,
max( decode( position, 4, column_name, null ) ) column_4,
max( decode( position, 5, decode(column_name, null, null, ',' ), null ) ) comma_5,
max( decode( position, 5, column_name, null ) ) column_5,
max( decode( position, 6, decode(column_name, null, null, ',' ), null ) ) comma_6,
max( decode( position, 6, column_name, null ) ) column_6,
max( decode( position, 7, decode(column_name, null, null, ',' ), null ) ) comma_7,
max( decode( position, 7, column_name, null ) ) column_7
from all_cons_columns
group by owner, table_name, constraint_name ) cc
where c.owner = 'BCC'
and c.generated != 'GENERATED NAME'
and cc.owner = c.owner
and cc.table_name = c.table_name
and cc.constraint_name = c.constraint_name
order by c.table_name,
decode(c.constraint_type,
'P','PK',
'R','FK',
'C','CHECK',
c.constraint_type) desc,
c.constraint_name;
begin
for c_columnRow in c_column loop
dbms_output.put_line(substr(c_columnRow.table_name||'.'||c_columnRow.column_name||': '||
c_columnRow.data_type||'('||
nvl(c_columnRow.data_precision, c_columnRow.data_length)||
c_columnRow.scale_comma||c_columnRow.data_scale||') '||
c_columnRow.default_equals||c_columnRow.data_default||
' <'||c_columnRow.nullable||'>',1,255));
rowcnt := rowcnt + 1;
end loop;
for c_constraintRow in c_constraint loop
dbms_output.put_line(substr(c_constraintRow.constraint_type||' '||c_constraintRow.constraint_name||' on '||
c_constraintRow.table_name||' ('||
c_constraintRow.search_condition||
c_constraintRow.r_columns||') ',1,255));
if length(c_constraintRow.constraint_type||' '||c_constraintRow.constraint_name||' on '||
c_constraintRow.table_name||' ('||
c_constraintRow.search_condition||
c_constraintRow.r_columns||') ') > 255 then
dbms_output.put_line('... '||substr(c_constraintRow.constraint_type||' '||c_constraintRow.constraint_name||' on '||
c_constraintRow.table_name||' ('||
c_constraintRow.search_condition||
c_constraintRow.r_columns||') ',256,251));
end if;
rowcnt := rowcnt + 1;
end loop;
end;
/
Unfortunately, there are a few limitations:
- Embedded carriage returns and whitespace in data_defaults, and check constraint definitions, may be highlighted as differences, even though they have zero effect on the schema.
- Does not include alternate keys, unique indexes or performance indexes. This would require a third SELECT statement in the script, referencing all_ind_columns and all_indexes catalog views.
- Does not include security details, synonyms, packages, triggers, etc. Packages and triggers would be best compared using an approach similar to the one you originally proposed. Other aspects of the schema definition could be added to the above script.
- The FK definitions above identify the referencing foreign key columns, but not the PK or the table being referenced. Just one more detail I never got around to doing.
Even if you don't use the script. There's a certain techie pleasure in playing with this stuff. ;-)
Matthew
回答2:
I'm afraid I can't do more for you at the moment, but this should give you a basic idea.
It selects ADD and DROP column statements that you could execute after carefully reviewing them.
It does not handle
- created/dropped tables
- data type/precision changes of existing columns (
ALTER TABLE MODIFY) DEFAULT VALUES(so you can't apply it on a table with data when new column isNOT NULL)- Check constraints, Foreign Key constraints
I tried it with some basic data-types (NUMBER, VARCHAR2, DATE) and it worked. Good luck :)
SELECT 'ALTER TABLE ' || LOWER(table_name)
|| ' ADD ' || LOWER(column_name) || ' ' || data_type
|| CASE WHEN data_type NOT IN ('DATE') THEN '(' || data_length || ')' END
|| CASE WHEN nullable='Y' THEN ' NOT NULL' END
|| ';' cmd
FROM all_tab_cols c2
WHERE owner = 'SCHEMA_1'
AND NOT EXISTS ( SELECT 1
FROM all_tab_cols c1
WHERE owner = 'SCHEMA_2'
AND c1.table_name = c2.table_name
AND c1.column_name = c2.column_name )
UNION ALL
SELECT 'ALTER TABLE ' || LOWER(table_name)
|| ' DROP COLUMN ' || LOWER(column_name) || ';'
FROM all_tab_cols c2
WHERE owner = 'SCHEMA_2'
AND NOT EXISTS ( SELECT 1
FROM all_tab_cols c1
WHERE owner = 'SCHEMA_1'
AND c1.table_name = c2.table_name
AND c1.column_name = c2.column_name )
ORDER BY cmd;
回答3:
I started writing an answer for this but my list of caveats became longer than the answer so I decided to scrap it.
You should go for a schema comparison tool.
There are free versions available - take a look at this question on Server Fault:
https://serverfault.com/questions/26360/how-can-i-diff-two-oracle-10g-schemas
My suggestion would be to download Oracle's SQL Developer and use the built-in schema diff tool (although this requires that you have the Change Management Pack license).
来源：https://stackoverflow.com/questions/2332617/compare-two-schemas-and-update-the-old-schema-with-the-new-columns-of-new-schema