Нужна помощь для настройки Oracle SQL-запроса

Question

В таблице temp есть столбцы «слово» и «предложение».Код ниже проверяет, есть ли в предложении какие-либо слова из словосочетания.Если слово существует, слово будет заменено на URL (содержит само слово и его идентификатор).Код работает нормально для 1-10 строк.Спасибо человеку, который помог мне добраться до этого здесь, в Stackoverflow.В таблице около 50 тыс. Записей.Он потребляет все временное пространство.Помогите, пожалуйста, рассмотреть и точно настроить запрос.

Требование: Есть 50 тыс. Слов и предложений.Требуется заменить слова в предложениях URL-адресом (содержит слово и его идентификатор), если какое-либо из слов существует в столбце слова.При поиске слов поиск должен быть без учета регистра.Кроме того, нам нужно сохранить тот же регистр в предложении при замене на URL.

Create table temp(
  id       NUMBER,
  word     VARCHAR2(1000),
  Sentence VARCHAR2(2000)
);

insert into temp
SELECT 1,'automation testing', 'automtestingation TeStInG TEST is popular kind of testing' FROM DUAL UNION ALL
SELECT 2,'testing','manual testing' FROM DUAL UNION ALL
select 2,'test', 'test' FROM DUAL UNION ALL
SELECT 3,'manual testing','this is an old method of testing' FROM DUAL UNION ALL
SELECT 4,'punctuation','automation testing,manual testing,punctuation,automanual testing-testing' FROM DUAL UNION ALL
SELECT 5,'B-number analysis','B-number analysis table' FROM DUAL UNION ALL
SELECT 6,'B-number analysis table','testing B-number analysis' FROM DUAL UNION ALL
SELECT 7,'Not Matched','testing testing testing' FROM DUAL

Типы SQL:

CREATE TYPE stringlist IS TABLE OF VARCHAR2(4000);
/
CREATE TYPE intlist IS TABLE OF NUMBER(20,0);
/

Функция PLSQL

CREATE FUNCTION replace_words(
  word_list IN  stringlist,
  id_list   IN  intlist,
  sentence  IN  temp.sentence%TYPE
) RETURN temp.sentence%TYPE
IS
  p_sentence       temp.sentence%TYPE := UPPER( sentence );
  p_pos            PLS_INTEGER := 1;
  p_min_word_index PLS_INTEGER;
  p_word_index     PLS_INTEGER;
  p_start          PLS_INTEGER;
  p_index          PLS_INTEGER;
  o_sentence       temp.sentence%TYPE;
BEGIN
  LOOP
    p_min_word_index := NULL;
    p_index          := NULL;
    FOR i IN 1 .. word_list.COUNT LOOP
      p_word_index := p_pos;
      LOOP
        p_word_index := INSTR( p_sentence, word_list(i), p_word_index );
        EXIT WHEN p_word_index = 0;
        IF (   p_word_index  > 1
           AND REGEXP_LIKE( SUBSTR( p_sentence, p_word_index - 1, 1 ), '\w' )
           )
           OR  REGEXP_LIKE( SUBSTR( p_sentence, p_word_index + LENGTH( word_list(i) ), 1 ), '\w' )
        THEN
           p_word_index := p_word_index + 1;
           CONTINUE;
        END IF;
        IF p_min_word_index IS NULL OR p_word_index < p_min_word_index THEN
          p_min_word_index := p_word_index;
          p_index := i;
        END IF;
        EXIT;
      END LOOP;
    END LOOP;
    IF p_index IS NULL THEN
      o_sentence := o_sentence || SUBSTR( sentence, p_pos );
      EXIT;
    ELSE
      o_sentence := o_sentence
                    || SUBSTR( sentence, p_pos, p_min_word_index - p_pos )
                    || 'http://localhost/'
                    || id_list(p_index)
                    || '/<u>'
                    || SUBSTR( sentence, p_min_word_index, LENGTH( word_list( p_index ) ) )
                    || '</u>';
      p_pos := p_min_word_index + LENGTH( word_list( p_index ) );
    END IF;
  END LOOP;
  RETURN o_sentence;
END;
/

Мердж

MERGE INTO temp dst
USING (
  WITH lists ( word_list, id_list ) AS (
    SELECT CAST(
             COLLECT(
               UPPER( word )
               ORDER BY LENGTH( word ) DESC, UPPER( word ) ASC, ROWNUM
             )
             AS stringlist
           ),
           CAST(
             COLLECT(
               id
               ORDER BY LENGTH( word ) DESC, UPPER( word ) ASC, ROWNUM
             )
             AS intlist
           )
    FROM   temp
  )
  SELECT t.ROWID rid,
         replace_words(
           word_list,
           id_list,
           sentence
         ) AS replaced_sentence
  FROM   temp t
         CROSS JOIN lists
) src
ON ( dst.ROWID = src.RID )
WHEN MATCHED THEN
  UPDATE SET sentence = src.replaced_sentence;

Answer 1

Я отделяю слова (с идентификаторами) от предложений и помещаю слова в нижний регистр, потому что вы все равно хотите поиск без учета регистра.Если я найду два совпадения в одной и той же позиции в предложении, я выберу более длинное.Если есть перекрытия («ручное тестирование» и «стратегия тестирования»), я всегда выбираю «слово», которое стоит первым в предложении.

С наилучшими пожеланиями, Стью Эштон

SQL> Create table temp(
  2    id       NUMBER,
  3    word     VARCHAR2(1000),
  4    Sentence VARCHAR2(2000)
  5  );

SQL> insert into temp
  2  SELECT 1,'automation testing', 'automtestingation TeStInG TEST is popular kind of testing' FROM DUAL UNION ALL
  3  SELECT 2,'testing','manual testing' FROM DUAL UNION ALL
  4  select 2,'test', 'test' FROM DUAL UNION ALL
  5  SELECT 3,'manual testing','this is an old method of testing' FROM DUAL UNION ALL
  6  SELECT 4,'punctuation','automation Testing,manual tEsting,punctuation,automanual teSting-tesTing' FROM DUAL UNION ALL
  7  SELECT 5,'B-number analysis','B-number analysis table' FROM DUAL UNION ALL
  8  SELECT 6,'B-number analysis table','testing B-number analysis' FROM DUAL UNION ALL
  9  SELECT 7,'Not Matched','Testing tEsting teSting' FROM DUAL;

SQL> create table sentences as select sentence from temp;

SQL> create table words cache as
  2  select length(word) word_length,
  3    min(id) id,
  4    lower(word) word
  5  from temp
  6  group by length(word), lower(word);

SQL> insert into sentences
  2  select listagg(word, ',') within group(order by word)
  3  from words;

SQL> insert into sentences values('Nothing matches here');

SQL> commit;

SQL> declare
  2    cursor cur_sentences is 
  3      select rowid rid, sentence from sentences s
  4      where exists (
  5        select null from words
  6        where instr(lower(s.sentence), word) > 0
  7      )
  8      for update;
  9    type tt_sentences is table of cur_sentences%rowtype;
 10    lt_sentences tt_sentences;
 11    lt_sentences_new tt_sentences;
 12  
 13    function change_sentence(p_sentence in sentences.sentence%type)
 14    return sentences.sentence%type is
 15      cursor cur_words(cp_sentence in sentences.sentence%type) is
 16        with recurse (pos, word_length, id, word) as (
 17          select regexp_instr(cp_sentence, '(^|\W)('||word||')(\W|$)', 1, 1, 0, 'i', 2),
 18            word_length, id, word
 19          from words
 20          where regexp_instr(cp_sentence, '(^|\W)('||word||')(\W|$)', 1, 1, 0, 'i', 2) > 0
 21          union all
 22          select regexp_instr(cp_sentence, '(^|\W)('||word||')(\W|$)', pos+1, 1, 0, 'i', 2),
 23            word_length, id, word
 24          from recurse
 25          where regexp_instr(cp_sentence, '(^|\W)('||word||')(\W|$)', pos+1, 1, 0, 'i', 2) > 0
 26        )
 27        select pos, word_length, id, word,
 28          substr(cp_sentence, pos, length(word)) new_word
 29        from recurse
 30        order by pos, word_length desc;
 31      type tt_words is table of cur_words%rowtype;
 32      lt_words tt_words;
 33      lt_words_kept tt_words:= new tt_words();
 34      l_pos number := 0;
 35      l_sentence sentences.sentence%type := p_sentence;
 36    begin
 37      open cur_words(p_sentence);
 38      fetch cur_words bulk collect into lt_words;
 39      for i in 1..lt_words.count loop
 40        if l_pos < lt_words(i).pos then
 41          l_pos := lt_words(i).pos + lt_words(i).word_length;
 42          lt_words_kept.extend;
 43          lt_words_kept(lt_words_kept.count) := lt_words(i);
 44        end if;
 45      end loop;
 46      close cur_words;
 47      for i in reverse 1..lt_words_kept.count loop
 48        l_sentence := regexp_replace(
 49          l_sentence,
 50          lt_words_kept(i).new_word,
 51          'http://localhost/'||lt_words_kept(i).id||'/<u>'||lt_words_kept(i).new_word||'</u>',
 52          lt_words_kept(i).pos,
 53          1
 54        );
 55      end loop;
 56      return l_sentence;
 57    exception when others then
 58      close cur_words;
 59      raise;
 60    end change_sentence;
 61  
 62  begin
 63    open cur_sentences;
 64    loop
 65      fetch cur_sentences bulk collect into lt_sentences limit 100;
 66      exit when lt_sentences.count = 0;
 67      lt_sentences_new := new tt_sentences();
 68      lt_sentences_new.extend(lt_sentences.count);
 69      for i in 1..lt_sentences.count loop
 70        lt_sentences_new(i).sentence := change_sentence(lt_sentences(i).sentence);
 71      end loop;
 72      forall i in 1..lt_sentences.count
 73        update sentences set sentence = lt_sentences_new(i).sentence where rowid = lt_sentences(i).rid;
 74      exit when cur_sentences%notfound;
 75    end loop;
 76    close cur_sentences;
 77  exception when others then
 78    if cur_sentences%isopen then
 79      close cur_sentences;
 80      raise;
 81    end if;
 82  end;
 83  /

PL/SQL procedure successfully completed.

SQL> select * from sentences order by 1;

SENTENCE                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                                        
--------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------------
Nothing matches here
automtestingation http://localhost/2/<u>TeStInG</u> http://localhost/2/<u>TEST</u> is popular kind of http://localhost/2/<u>testing</u>
http://localhost/1/<u>automation Testing</u>,http://localhost/3/<u>manual tEsting</u>,http://localhost/4/<u>punctuation</u>,automanual http://localhost/2/<u>teSting</u>-http://localhost/2/<u>tesTing</u>
http://localhost/1/<u>automation testing</u>,http://localhost/5/<u>b-number analysis</u>,http://localhost/6/<u>b-number analysis table</u>,http://localhost/3/<u>manual testing</u>,http://localhost/7/<u>not matched</u>,http://localhost/4/<u>punctuation</u>,http://localhost/2/<u>test</u>,http://localhost/2/<u>testing</u>
http://localhost/2/<u>Testing</u> http://localhost/2/<u>tEsting</u> http://localhost/2/<u>teSting</u>
http://localhost/2/<u>test</u>
http://localhost/2/<u>testing</u> http://localhost/5/<u>B-number analysis</u>
http://localhost/3/<u>manual testing</u>
http://localhost/6/<u>B-number analysis table</u>
this is an old method of http://localhost/2/<u>testing</u>