Versions Compared

Key

  • This line was added.
  • This line was removed.
  • Formatting was changed.

...

The new feature was implemented to make it possible to browse call-numbers by their types. It was implemented by using references to call-number types and changing logic for calculating shelving orderĀ that is used for call-number sorting.

Instructions

Step 1

Release info:

Required: from Orchid to Poppy/CSP1/CSP2/CSP3

Not required: from Poppy to CSP1/CSP2/CSP3, from CSP1 to CSP2/CSP3, from CSP2 to CSP3

Run time:

Before upgrade

Precondition:

...

Step NExecution time

1

1s
21h 10m
320m
41h 30m
51s
62s
Total3h

Step 2

...

Precondition

Run time:

After upgrade, before reindex

PreconditionRelease info:

...

Requires upgrade to Poppy CSP3 before running scripts

Action:

Run script:

Code Block
languagesql
titleRecalculate shelving order
collapsetrue
Update sql functions
-- Change search path for the script
SET search_path = <tenant>_mod_inventory_storage;

-- Crete/replace function to recalculate effective values and shelving order in items
CREATE OR REPLACE FUNCTION set_item_effective_values(holding jsonb, item jsonbCREATE OR REPLACE FUNCTION dewey_call_number(call_number text)
  RETURNS jsonbtext AS $$
  BEGINDECLARE
    RETURN set_effective_shelving_order(matches text[];
    class_digits text;
    itemclass_decimal text;
    cutter text;
   || jsonb_build_object('effectiveLocationId',
 other text;
  BEGIN
    matches = regexp_match(call_number,     COALESCE(item->>'temporaryLocationId', item->>'permanentLocationId', holding->>'effectiveLocationId'))'^(\d+)(\.\d+)? *\.?(?:([A-Z]\d{1,3}(?:[A-Z]+)?) *(.*)|(.*))$');
    IF matches IS NULL THEN
      ||RETURN effectivenumerically_sortable(trim2null(call_number_components(holding, item));
    END IF;
$$ LANGUAGE plpgsql IMMUTABLE PARALLELclass_digits SAFE STRICT;

-- Recalculate effective values and shelving order in items
UPDATE item
SET jsonb = set_item_effective_values(item_data.holdings, item_data.item)
FROM (
    select item.id, hr.jsonb as holdings, item.jsonb as item from holdings_record hr
    join item on item.holdingsRecordId = hr.id
    where trim2null(hr.jsonb->>'callNumberTypeId') IS NOT NULL
    and trim2null(item.jsonb->'effectiveCallNumberComponents'->>'typeId') IS NULL
) AS item_data
WHERE item.id = item_data.id;

Action timing:

Tested on database with 9M rows in item table.

Execution time: TBD

...

= matches[1];
    class_decimal = matches[2];
    cutter = matches[3];
    other = numerically_sortable(trim2null(concat(trim2null(matches[4]), trim2null(matches[5]))));
    RETURN concat_ws(' ', concat(sortable_number(class_digits), class_decimal),
                          cutter,
                          other
                    );
  END;
$$ LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;

CREATE OR REPLACE FUNCTION lc_nlm_call_number(call_number text)
  RETURNS text AS $$
  DECLARE
    matches text[];
    classification text;
    classLetters text;
    classDigits text;
    classDecimal text;
    everythingElse text;
    classSuffix text;
    cutter text;
  BEGIN
    call_number = upper(call_number);
    matches = regexp_match(call_number, '^(([A-Z]+) *(?:(\d+)(\.\d+)?)?)(.*)$');
    IF matches IS NULL THEN
      RETURN trim(cutter_shelf_key(trim2null(call_number)));
    END IF;
    classification = trim(matches[1]);
    classLetters = trim(matches[2]);
    classDigits = trim(matches[3]);
    classDecimal = trim(matches[4]);
    everythingElse = matches[5];
    IF classDigits IS NULL THEN
      RETURN NULL;
    END IF;
    IF length(everythingElse) > 0 THEN
      -- combining greedy and non-greedy:
      -- https://www.postgresql.org/docs/current/functions-matching.html#POSIX-MATCHING-RULES
      matches = regexp_match(everythingElse, '(?:(.*?)(\.?[A-Z]\d+|^\.[A-Z]| \.[A-Z])(.*)){1,1}');
      IF matches IS NULL THEN
        classSuffix = trim2null(everythingElse);
      ELSE
        classSuffix = trim2null(matches[1]);
        cutter = trim(matches[2] || matches[3]);
      END IF;
    END IF;
    classSuffix = numerically_sortable(classSuffix);
    IF substr(classSuffix, 1, 1) BETWEEN 'A' AND 'Z' THEN
      classSuffix = '_' || classSuffix;
    END IF;
    cutter = cutter_shelf_key(cutter);
    return trim(concat_ws(' ', classLetters,
                               concat(length(classDigits), classDigits, classDecimal),
                               trim2null(classSuffix),
                               trim(cutter)
                         ));
  END;
$$ LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;

CREATE OR REPLACE FUNCTION su_doc_call_number(call_number text)
  RETURNS text AS $$
  DECLARE
    matches text[];
  BEGIN
    matches = regexp_match(upper(call_number),
        '^([A-Z]+)\s*(\d+)(\.(?:[A-Z]+\d*|\d+))(/(?:[A-Z]+(?:\d+(?:-\d+)?)?|\d+(?:-\d+)?))?:?(.*)$');
    IF matches IS NULL THEN
      RETURN su_doc_part(trim2null(upper(call_number)));
    END IF;
    RETURN concat_ws(' ', matches[1], su_doc_part(matches[2]), su_doc_part(matches[3]), su_doc_part(matches[4]), su_doc_part(matches[5]));
  END;
$$ LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;

CREATE OR REPLACE FUNCTION sortable_number(n text)
  RETURNS text AS $$
  DECLARE int_part text;
  BEGIN
    n = regexp_replace(n, '^0+', '');
    int_part = split_part(n, '.', 1);
    RETURN concat(length(int_part), n);
  END;
$$ LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;

CREATE OR REPLACE FUNCTION cutter_shelf_key(s text)
  RETURNS text AS $$
  DECLARE
    chunk text;
    matches text[];
    cutter text;
    suffix text;
    result text;
  BEGIN
    FOREACH chunk IN ARRAY regexp_split_to_array(s, '(?=\.?[A-Z][0-9])') LOOP
      matches = regexp_match(chunk, '([A-Z][0-9]+)(.*)');
      IF matches IS NULL THEN
        -- before the first cutter
        IF result IS NULL THEN
            result = trim2null(numerically_sortable(chunk));
        ELSE
            result = concat_ws(' ', result, trim2null(numerically_sortable(chunk)));
        END IF;
      ELSE
        cutter = matches[1];
        suffix = trim2null(numerically_sortable(matches[2]));
        result = concat_ws(' ', result, cutter, suffix);
      END IF;
    END LOOP;
    RETURN result;
  END;
$$ LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;

Step 2 (Option 1)

Release info:

Requires upgrade to Poppy CSP3 before running scripts

Run time:

After upgrade, before reindex

Precondition:

No preconditions

Action:

...

Run script:

Code Block
languagesql
titleRecalculate shelving order
collapse

...

true
-- Change search path for the script
SET search_path = <tenant>_mod_inventory_storage

...

;

...


-- 

...

Recalculate effective values and shelving order in items
UPDATE item
SET jsonb = set_item_effective_values(item_data.holdings, item_data.item)
FROM (
    select item.id, hr.jsonb as holdings, item.jsonb as item from holdings_record hr
    join item on item.holdingsRecordId = hr.id
    where item.jsonb->'effectiveCallNumberComponents'->'callNumber' IS NOT NULL
    and item.jsonb->'effectiveCallNumberComponents'->'callNumber' != 'null'
) AS item_data
WHERE item.id = item_data.id;

Action timing:

Tested on database with 9M rows in item table.

Execution time: TBD


Step 2 (Option 2)

Release info:

Requires upgrade to Poppy CSP3 before running scripts

Run time:

After upgrade, before reindex

Precondition:

No preconditions

Action:

  1. Disable triggers

    Code Block
    languagesql
    collapsetrue
    createALTER table public.uuid_range
    (
        id             serial primary key,
        partition      uuid,
        subrange_start uuid,
        subrange_end   uuid,
        completed      boolean default false
    );-- If table exists and filled then do:
    UPDATE public.uuid_range SET completed=false;
    -- If table was just created then fill it:
    
    DO
    $$
        DECLARE
            partition_start UUID;
            subrange_start  UUID;
            subrange_end    UUID;
        BEGIN
            FOR i IN 0..15
                LOOP
                    partition_start := (rpad(to_hex(i), 8, '0') || '-0000-0000-0000-000000000000')::UUID;
                    subrange_start := partition_start;
    
                    FOR j IN 1..4096
           TABLE {tenant}_mod_inventory_storage.item DISABLE TRIGGER audit_item;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER check_statistical_code_references_on_insert;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER check_statistical_code_references_on_update;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER set_id_in_jsonb;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER set_item_md_json_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER set_item_md_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER set_item_ol_version_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER update_item_references;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER update_item_status_date;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER updatecompleteupdateddate_item_delete;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item DISABLE  TRIGGER updatecompleteupdateddate_item_insert_update;


  2. Create uuid_range table

    Code Block
    languagesql
    collapsetrue
    create table public.uuid_range
    (
        id             serial primary key,
        partition      uuid,
        subrange_start uuid,
        subrange_end   uuid,
        completed      boolean default false
    );-- If table exists and filled then do:
    UPDATE public.uuid_range SET completed=false;
    -- If table was just created then fill it:
    
    DO
    $$
        DECLARE
            partition_start UUID;
            subrange_start  UUID;
     LOOP       subrange_end    UUID;
        BEGIN
            IFFOR i <IN 0..15
         OR (i = 15 AND j < 4096)LOOP
    THEN                partition_start := (rpad(to_hex(i), 8, '0') || '-0000-0000-0000-000000000000')::UUID;
          if (j < 4096) then      subrange_start := partition_start;
    
                    FOR j IN 1..4096
       subrange_end := (to_hex(i)::text || rpad(lpad(to_hex(j), 3, '0'), 7, '0') ||         LOOP
                            IF i < 15 OR (i = 15 AND j < 4096) THEN
       '-0000-0000-0000-000000000000')::UUID;                         if (j < 4096) elsethen
                                    subrange_end := (to_hex(i)::text || rpad(lpad(to_hex(i + 1j), 3, '0'), 87, '0') || '-0000-0000-0000-000000000000')::UUID;
                                          end if;          '-0000-0000-0000-000000000000')::UUID;
                  ELSE              else
                  subrange_end := 'ffffffff-ffff-ffff-ffff-ffffffffffff'::UUID; -- upper bound for last subrange in last partition                 subrange_end := (rpad(to_hex(i + 1), 8, '0') || '-0000-0000-0000-000000000000')::UUID;
              END IF;                 end if;
           INSERT INTO public.uuid_range (partition, subrange_start, subrange_end)            ELSE
                VALUES (partition_start, subrange_start, subrange_end);             subrange_end := 'ffffffff-ffff-ffff-ffff-ffffffffffff'::UUID; -- upper bound for last subrange in last partition
     subrange_start := subrange_end;                     END LOOP;IF;
    
                      END LOOP;     END;INSERT $$
    LANGUAGE plpgsql; 

    Create/replace function to recalculate effective values and shelving order in items

    Code Block
    languagesql
    SET search_path = <tenant>_mod_inventory_storage;
    
    CREATE OR REPLACE FUNCTION set_item_effective_values(holding jsonb, item jsonb)
      RETURNS jsonb AS $$
      BEGININTO public.uuid_range (partition, subrange_start, subrange_end)
                            RETURNVALUES set_effective_shelving_order((partition_start, subrange_start, subrange_end);
    
                item           || jsonb_build_object('effectiveLocationId',subrange_start := subrange_end;
                       COALESCE(item->>'temporaryLocationId', item->>'permanentLocationId', holding->>'effectiveLocationId')) END LOOP;
               || effective_call_number_components(holding, item)) END LOOP;
        END;
    $$
    LANGUAGE plpgsql IMMUTABLE PARALLEL SAFE STRICT;; 


  3. Update step should be executed for each partition created from step #2. Update script can be executed concurrently.

    Code Block
    languagesql
    collapsetrue
    set search_path = "{tenant}_mod_inventory_storage", "public";
    
    do
    $$
        declare
            partition_to_process uuid := '{partition}'::uuid;
            range_rec           public.uuid_range%rowtype;
        begin
            for range_rec in select ur.*
                             from public.uuid_range ur
                             where ur.partition = partition_to_process
                               and ur.completed = false
                loop
                    --             start transaction;
                    UPDATE item i SET jsonb = set_item_effective_values(item_data.holdings, item_data.item)
    				FROM (
    					select item.id, hr.jsonb as holdings, item.jsonb as item from holdings_record hr
    					join item on item.holdingsRecordId = hr.id
                    	where item.id between range_rec.subrange_start and range_rec.subrange_end
                    	and item.jsonb->'effectiveCallNumberComponents'->'callNumber' IS NOT NULL
    					and item.jsonb->'effectiveCallNumberComponents'->'callNumber' != 'null'
    				) AS item_data
    				WHERE i.id = item_data.id;
     
                    update uuid_range
                    set completed = true
                    where id = range_rec.id;
                    --
                    raise notice 'subrange id: % completed', range_rec.id;
                    commit;
                end loop;
        end
    $$;


  4. Monitor how many sub-partitions in partitions are left to be processed

    Code Block
    languagesql
    collapsetrue
    select partition, count(partition)
    from public.uuid_range
    where completed = false
    group by partition
    order by partition;  


  5. Enable triggers

    Code Block
    languagesql
    collapsetrue
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER audit_item;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER check_statistical_code_references_on_insert;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER check_statistical_code_references_on_update;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER set_id_in_jsonb;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER set_item_md_json_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER set_item_md_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER set_item_ol_version_trigger;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER update_item_references;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER update_item_status_date;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER updatecompleteupdateddate_item_delete;
    
    ALTER TABLE {tenant}_mod_inventory_storage.item ENABLE TRIGGER updatecompleteupdateddate_item_insert_update;


Step 2 (Option 3). Requires Poppy CSP #3. Unstable on big amounts of data

Run time:

After upgrade, before reindex

...

Code Block
languagesql
-- Change search path for the script
SET search_path = <tenant>_mod_inventory_storage;

-- Recalculate effective values for items with no call number type specified. No shelving order recalculation
UPDATE item
SET jsonb = set_item_effective_values_no_shelv_recalc(item_data.holdings, item_data.item)
FROM (
    select item.id, hr.jsonb as holdings, item.jsonb as item from holdings_record hr
    join item on item.holdingsRecordId = hr.id
    where trim2null(hr.jsonb->>'callNumberTypeId') IS NOT NULL
    and trim2null(item.jsonb->'effectiveCallNumberComponents'->>'typeId') IS NULL
) AS item_data
WHERE item.id = item_data.id;

Step 3. Requires Poppy CSP #3. Unstable on big amounts of data

  1. Run async item shelving order migration via api. Anytime, may be during runtime. Values surrounded by "<>" to be replaced

    Code Block
    languagetext
    curl --location --request POST '<okapi-url>/inventory-storage/migrations/jobs' \
    --header 'X-Okapi-Tenant: <tenant>' \
    --header 'X-Okapi-Token: <token>' \
    --header 'Content-Type: application/json' \
    --data '{
      "migrations": ["itemShelvingOrderMigration"],
      "affectedEntities": ["ITEM"]
    }'


  2. Monitor execution until response body parameter "jobStatus" is "Completed". <jobId> is returned from previous request as "id" parameter in response body

    Code Block
    languagetext
    curl --location --request GET '<okapi-url>/inventory-storage/migrations/jobs/<jobId>' \
    --header 'X-Okapi-Tenant: <tenant>' \
    --header 'X-Okapi-Token: <token>' \
    --header 'Content-Type: application/json'


  3. Once migration is completed - reindex instances

...