Migrating from a legacy system


When you migrate to Evergreen, you generally want to migrate the bibliographic records and item information that existed in your previous library system. For anything more than a few thousand records, you should import the data directly into the database rather than use the tools in the staff client. While the data that you can extract from your legacy system varies widely, this section assumes that you or members of your team have the ability to write scripts and are comfortable working with SQL to manipulate data within PostgreSQL. If so, then the following section will guide you towards a method of generating common data formats so that you can then load the data into the database in bulk.

Making electronic resources visible in the catalog

Electronic resources generally do not have any call number or item information associated with them, and Evergreen enables you to easily make bibliographic records visible in the public catalog within sections of the organizational unit hierarchy. For example, you can make a set of bibliographic records visible only to specific branches that have purchased licenses for the corresponding resources, or you can make records representing publicly available electronic resources visible to the entire consortium.

Therefore, to make a record visible in the public catalog, modify the records using your preferred MARC editing approach to ensure the 856 field contains the following information before loading records for electronic resources into Evergreen:

Table 1. 856 field for electronic resources: indicators and subfields
Attribute Value Note

Indicator 1


Indicator 2

0 or 1

Subfield u

URL for the electronic resource

Subfield y

Text content of the link

Subfield z

Public note

Normally displayed after the link

Subfield 9

Organizational unit short name

The record will be visible when a search is performed specifying this organizational unit or one of its children. You can repeat this subfield as many times as you need.

Once your electronic resource bibliographic records have the required indicators and subfields for each 856 field in the record, you can proceed to load the records using either the command-line bulk import method or the MARC Batch Importer in the staff client.

Migrating your bibliographic records

Convert your MARC21 binary records into the MARCXML format, with one record per line. You can use the following Python script to achieve this goal; just install the pymarc library first, and adjust the values of the input and output variables as needed.

#!/usr/bin/env python
# -*- coding: utf-8 -*-
import codecs
import pymarc

input = 'records_in.mrc'
output = 'records_out.xml'

reader = pymarc.MARCReader(open(input, 'rb'), to_unicode=True)
writer = codecs.open(output, 'w', 'utf-8')
for record in reader:
    record.leader = record.leader[:9] + 'a' + record.leader[10:]
    writer.write(pymarc.record_to_xml(record) + "\n")

Note: pymarc does not fail gracefully when it encounters encoding issues - often either because LDR09 is lying, or because non-UTF8 / non-MARC8 chars (like iso-8859-1) have been added by other less strict tools and the above script will fail.

Another way to convert the .mrc to xml is with yaz-marcdump using the following:

yaz-marcdump -i marc -o marcxml -f utf8 -t utf8 -l 9=97 records_in.mrc > records_out.xml

The resulting file will have its XML formatted into human readable format and need to be massaged such that each record container is on a single line.

Once you have a MARCXML file with one record per line, you can load the records into your Evergreen system via a staging table in your database.

  1. Connect to the PostgreSQL database using the psql command. For example:

    psql -U <user-name> -h <hostname> -d <database>
  2. Create a staging table in the database. The staging table is a temporary location for the raw data that you will load into the production table or tables. Issue the following SQL statement from the psql command line, adjusting the name of the table from staging_records_import, if desired:

    CREATE TABLE staging_records_import (id BIGSERIAL, dest BIGINT, marc TEXT);
  3. Create a function that will insert the new records into the production table and update the dest column of the staging table. Adjust "staging_records_import" to match the name of the staging table that you plan to create when you issue the following SQL statement:

    FOR stage IN SELECT * FROM staging_records_import ORDER BY id LOOP
          INSERT INTO biblio.record_entry (marc, last_xact_id) VALUES (stage.marc, 'IMPORT');
          UPDATE staging_records_import SET dest = currval('biblio.record_entry_id_seq')
           WHERE id = stage.id;
       END LOOP;
      $$ LANGUAGE plpgsql;
  4. Load the data from your MARCXML file into the staging table using the COPY statement, adjusting for the name of the staging table and the location of your MARCXML file:

    COPY staging_records_import (marc) FROM '/tmp/records_out.xml';
  5. Load the data from your staging table into the production table by invoking your staging function:

    SELECT staging_importer();

When you leave out the id value for a BIGSERIAL column, the value in the column automatically increments for each new record that you add to the table.

Once you have loaded the records into your Evergreen system, you can search for some known records using the staff client to confirm that the import was successful.

Migrating your call numbers, items, and parts

'Holdings', comprised of call numbers, items, and parts, are the set of objects that enable users to locate and potentially acquire materials from your library system.

'Call numbers' connect libraries to bibliographic records. Each call number has a 'label' associated with a classification scheme such as a the Library of Congress or Dewey Decimal systems, and can optionally have either or both a label prefix and a label suffix. Label prefixes and suffixes do not affect the sort order of the label.

'Copies' connect call numbers to particular instances of that resource at a particular library. Each item has a barcode and must exist in a particular item location. Other optional attributes of items include circulation modifier, which may affect whether that item can circulate or for how long it can circulate, and OPAC visibility, which controls whether that particular item should be visible in the public catalog.

'Parts' provide more granularity for items, primarily to enable patrons to place holds on individual parts of a set of items. For example, an encyclopedia might be represented by a single bibliographic record, with a single call number representing the label for that encyclopedia at a given library, with 26 items representing each letter of the alphabet, with each item mapped to a different part such as A, B, C, …​ Z.

To migrate this data into your Evergreen system, you will create another staging table in the database to hold the raw data for your materials from which the actual call numbers, items, and parts will be generated.

Begin by connecting to the PostgreSQL database using the psql command. For example:

psql -U <user-name> -h <hostname> -d <database>

Create the staging materials table by issuing the following SQL statement:

CREATE TABLE staging_materials (
  bibkey BIGINT,  -- biblio.record_entry_id
  callnum TEXT, -- call number label
  callnum_prefix TEXT, -- call number prefix
  callnum_suffix TEXT, -- call number suffix
  callnum_class TEXT, -- classification scheme
  create_date DATE,
  location TEXT, -- shelving location code
  item_type TEXT, -- circulation modifier code
  owning_lib TEXT, -- org unit code
  barcode TEXT, -- copy barcode
  part TEXT

For the purposes of this example migration of call numbers, items, and parts, we assume that you are able to create a tab-delimited file containing values that map to the staging table properties, with one item per line. For example, the following 5 lines demonstrate how the file could look for 5 different items, with non-applicable attribute values represented by \N, and 3 of the items connected to a single call number and bibliographic record via parts:

1   QA 76.76 A3 \N  \N  LC  2012-12-05  STACKS  BOOK    BR1 30007001122620  \N
2   GV 161 V8   Ref.    Juv.    LC  2010-11-11  KIDS    DVD BR2 30007005197073  \N
3   AE 5 E363 1984  \N  \N      LC  1984-01-10  REFERENCE   BOOK    BR1 30007006853385  A
3   AE 5 E363 1984  \N  \N      LC  1984-01-10  REFERENCE   BOOK    BR1 30007006853393  B
3   AE 5 E363 1984  \N  \N      LC  1984-01-10  REFERENCE   BOOK    BR1 30007006853344  C

Once your holdings are in a tab-delimited format—​which, for the purposes of this example, we will name holdings.tsv--you can import the holdings file into your staging table. Copy the contents of the holdings file into the staging table using the COPY SQL statement:

COPY staging_items (bibkey, callnum, callnum_prefix,
  callnum_suffix, callnum_class, create_date, location,
  item_type, owning_lib, barcode, part) FROM 'holdings.tsv';

Generate the item locations you need to represent your holdings:

INSERT INTO asset.copy_location (name, owning_lib)
  SELECT DISTINCT location, 1 FROM staging_materials
    SELECT 1 FROM asset.copy_location
    WHERE name = location

Generate the circulation modifiers you need to represent your holdings:

INSERT INTO config.circ_modifier (code, name, description, sip2_media_type)
  SELECT DISTINCT circmod, circmod, circmod, '001'
  FROM staging_materials
    SELECT 1 FROM config.circ_modifier
    WHERE circmod = code

Generate the call number prefixes and suffixes you need to represent your holdings:

INSERT INTO asset.call_number_prefix (owning_lib, label)
  SELECT DISTINCT aou.id, callnum_prefix
  FROM staging_materials sm
    INNER JOIN actor.org_unit aou
      ON aou.shortname = sm.owning_lib
    SELECT 1 FROM asset.call_number_prefix acnp
    WHERE callnum_prefix = acnp.label
      AND aou.id = acnp.owning_lib
  ) AND callnum_prefix IS NOT NULL;

INSERT INTO asset.call_number_suffix (owning_lib, label)
  SELECT DISTINCT aou.id, callnum_suffix
  FROM staging_materials sm
    INNER JOIN actor.org_unit aou
      ON aou.shortname = sm.owning_lib
    SELECT 1 FROM asset.call_number_suffix acns
    WHERE callnum_suffix = acns.label
      AND aou.id = acns.owning_lib
  ) AND callnum_suffix IS NOT NULL;

Generate the call numbers for your holdings:

INSERT INTO asset.call_number (
  creator, editor, record, owning_lib, label, prefix, suffix, label_class
  SELECT DISTINCT 1, 1, bibkey, aou.id, callnum, acnp.id, acns.id,
  CASE WHEN callnum_class = 'LC' THEN 1
             WHEN callnum_class = 'DEWEY' THEN 2
  FROM staging_materials sm
    INNER JOIN actor.org_unit aou
      ON aou.shortname = owning_lib
    INNER JOIN asset.call_number_prefix acnp
      ON COALESCE(acnp.label, '') = COALESCE(callnum_prefix, '')
    INNER JOIN asset.call_number_suffix acns
      ON COALESCE(acns.label, '') = COALESCE(callnum_suffix, '')

Generate the items for your holdings:

INSERT INTO asset.copy (
  circ_lib, creator, editor, call_number, location,
 loan_duration, fine_level, barcode
  SELECT DISTINCT aou.id, 1, 1, acn.id, acl.id, 2, 2, barcode
  FROM staging_materials sm
    INNER JOIN actor.org_unit aou
      ON aou.shortname = sm.owning_lib
    INNER JOIN asset.copy_location acl
      ON acl.name = sm.location
    INNER JOIN asset.call_number acn
      ON acn.label = sm.callnum
  WHERE acn.deleted IS FALSE

Generate the parts for your holdings. First, create the set of parts that are required for each record based on your staging materials table:

INSERT INTO biblio.monograph_part (record, label)
  SELECT DISTINCT bibkey, part
  FROM staging_materials sm
    SELECT 1 FROM biblio.monograph_part bmp
    WHERE sm.part = bmp.label
      AND sm.bibkey = bmp.record

Now map the parts for each record to the specific items that you added:

INSERT INTO asset.copy_part_map (target_copy, part)
  SELECT DISTINCT acp.id, bmp.id
  FROM staging_materials sm
    INNER JOIN asset.copy acp
      ON acp.barcode = sm.barcode
    INNER JOIN biblio.monograph_part bmp
      ON bmp.record = sm.bibkey
    AND part = bmp.label
    AND acp.deleted IS FALSE
    SELECT 1 FROM asset.copy_part_map
    WHERE target_copy = acp.id
      AND part = bmp.id

At this point, you have loaded your bibliographic records, call numbers, call number prefixes and suffixes, items, and parts, and your records should be visible to searches in the public catalog within the appropriate organization unit scope.