Instructions for using sqlite DDLs for non-R users

未关闭
#693 4 条评论 5 个 reaction 已指派 0 人 在 GitHub 查看

还没有人认领这个 Issue。

评估

难度
4/5
预计耗时
3-5 天
新手友好度
35/100
Issue 类型
文档
描述清晰度
基本清楚
活跃度
停滞
技术栈
bash, sql, sqlite

调研方向

检查提供的 update-ddl.sh 脚本以及 OMOPCDM*_ddl.sql、_primary_keys.sql、_constraints.sql 和 *_indices.sql 文件;检查 sqlite3 如何处理这些文件,以及 sqlrender 如何使用 @cdmDatabaseSchema。记录一个可重复的 non-R 工作流,并验证该工作流能够创建数据库且不会出现 foreign-key 错误。

由索引模型根据 Issue 内容生成。

描述

Bug Community Contribution

Hi there!

I've been playing around trying to get the DDLs loaded for sqlite3, version 5.4 using the sqlite3 command line and I've been having a pretty rough go of it using the provided definitions.

I didn't find any documentation here or elsewhere about the @cdmDatabaseSchema notation in the ddl files, so I went ahead and removed them. I assume they are something used by the sqlrender utility - but I'm not an R user - I'm just trying to create a working OMOPCDM database in sqlite.

I also discovered that sqlite does not allow you to apply primary or foreign keys to tables using the ALTER TABLE commands provided. So, I wrote a brittle little script to apply the PK and FK definitions to the main _ddl.sql file.

This allows me to use the patched _ddl.sql file to define tables with PKs and FKs. The _indices.sql file seems fine as-is.

I did notice that the COHORT table is missing a PRIMARY KEY, so the foreign key check fails in a clean install for just that table. It almost seems like COHORT should have cohort_id integer PRIMARY KEY and COHORT_DEFINITION should have cohort_definition_id integer PRIMARY KEY - and then the COHORT table should have a FK on COHORT_DEFINITION.cohort_definition_id (instead, it seems backwards). But as it is now, if I make COHORT.cohort_definition_id the PK of the COHORT table - it gets me past the failing foreign key checks. Anyway, I expect some of that will change in future versions - my script includes a hack for this.

For posterity, I'm providing my script here. However, are there instructions for people who don't want to use R but just want to create a working DB out there somewhere?

Click to see my script
#!/bin/bash
#
# File: update-ddl.sh
#
# Converts the OMOP-provided DDL database definition scripts into a format
# that can be loaded by sqlite3.
#
# This has been tested against the provided sqlite 5.4 DDL files.
#
# To run this, place copies of the provided .sql files into this directory
# and run the script.  It will modify the sql files and create an empty
# database called cdm.db
#

set -e
set -o pipefail
set -u


# There should be only one DDL file in the local directory.
DDL=$( ls OMOPCDM*_ddl.sql )
[ ! -e ${DDL} ] && echo "Failed to find an OMOPCDM*_ddl.sql schema!" && exit 1


# Awk scriptlet to add a PRIMARY KEY declaration to a table.column in a table
# definition.
#
# Assumptions:
#   'table' is padded with spaces
#   'column' begins with a tab and ends with a space
#
AWK_PK='
/^CREATE TABLE / {
  if ($0 ~ table) { target=1 } else { target=0 }
}

{
  if (target && $0 ~ column && (!($0 ~ / PRIMARY KEY,/))) {
    sub(/,/, " PRIMARY KEY,");
  }
  print $0
}
'

# Awk scriptlet to add a FOREIGN KEY declaration to a table.column in a table
# definition.
#
# Assumptions:
#   'table' is padded with spaces
#   'column' begins with a tab and ends with a space
#   'ref' is multiple words long
#
AWK_FK='
/^CREATE TABLE / {
  if ($0 ~ table) { target=1 } else { target=0 }
}

{
  if (target && $0 ~ column && (!($0 ~ / REFERENCES /))) {
    if ($0 ~ /,$/) {
      sub(/,$/, " REFERENCES " ref ",");
    } else {
      sub(/ );$/, " REFERENCES " ref ");");
    }
  }
  print $0
}
'

##
# Gets the table names and PK columns from constraints.
#
function get_pks() {
  grep '^ALTER TABLE ' OMOPCDM*_primary_keys.sql \
    | sed -e 's:.* \(.*\) ADD CONSTRAINT .* PRIMARY KEY .\(.*\).;:\1 \2:'
}


##
# Decorate the PK columns with PRIMARY KEY if not already decorated.
#
function mark_pk() {

  # Apply the change to the ddl file content, saving into a temp file.
  cat ${DDL} | awk -v table=" ${1} " -v column="        ${2} " "${AWK_PK}" \
    > ${DDL}.tmp

  # diff the files to assert diff count is 1 changed line only.
  (( $( diff -by --suppress-common-lines ${DDL} ${DDL}.tmp | wc -l ) == 1 )) || exit 1

  # Apply the temp file to the DDL file.
  mv ${DDL}.tmp ${DDL}
}


function get_fks() {
  grep '^ALTER TABLE ' OMOPCDM*_constraints.sql \
    | sed -e 's:ALTER TABLE \(.*\) ADD CONSTRAINT .* FOREIGN KEY .\([^ ]*\). REFERENCES:\1 \2:' \
    | sed -e 's:;::'
}


##
# Decorate the PK columns with PRIMARY KEY if not already decorated.
#
function mark_fk() {

  # Apply the change to the ddl file content, saving into a temp file.
  cat ${DDL} \
    | awk -v table=" ${1} " -v column=" ${2} " -v ref="${3}" "${AWK_FK}" \
    > ${DDL}.tmp

  # diff the files to assert diff count is 1 changed line only.
  (( $( diff -by --suppress-common-lines ${DDL} ${DDL}.tmp | wc -l ) == 1 )) || exit 1

  # Apply the temp file to the DDL file.
  mv ${DDL}.tmp ${DDL}
}


# Remove the template schema reference, which isn't supported in sqlite.
sed -i -e 's:@cdmDatabaseSchema.::g' OMOPCDM*.sql && rm -f OMOPCDM*.sql-e

# Update the table definitions to insert the PRIMARY KEY declarations.
echo "Adding PRIMARY KEYS to DDL"
get_pks | while read table column; do
  mark_pk ${table} ${column}
done

# Insert the FK constraints into the table.
echo "Adding FOREIGN KEYS to DDL"
get_fks | while read table column reference; do
  mark_fk ${table} ${column} "${reference}"
done

# HACK: no PK in the cohort table, so this is needed to satisfy FK checks.
mark_pk cohort cohort_definition_id

# Populate the DB tables and indices.
echo "Creating empty cdm.db database"
rm -f cdm.db
touch cdm.db
sqlite3 cdm.db < ${DDL}
sqlite3 cdm.db < OMOPCDM*_indices.sql
sqlite3 cdm.db "pragma foreign_key_check"

# Success!
echo "DONE!"

This updates the _ddl.sql file to look like this:

CREATE TABLE person (
                        person_id integer NOT NULL PRIMARY KEY,
                        gender_concept_id integer NOT NULL REFERENCES CONCEPT (CONCEPT_ID),
                        year_of_birth integer NOT NULL,
                        month_of_birth integer NULL,
                        day_of_birth integer NULL,
                        birth_datetime REAL NULL,
                        race_concept_id integer NOT NULL REFERENCES CONCEPT (CONCEPT_ID),
                        ethnicity_concept_id integer NOT NULL REFERENCES CONCEPT (CONCEPT_ID),
                        location_id integer NULL REFERENCES LOCATION (LOCATION_ID),
                        provider_id integer NULL REFERENCES PROVIDER (PROVIDER_ID),
                        care_site_id integer NULL REFERENCES CARE_SITE (CARE_SITE_ID),
                        person_source_value TEXT NULL,
                        gender_source_value TEXT NULL,
                        gender_source_concept_id integer NULL REFERENCES CONCEPT (CONCEPT_ID),
                        race_source_value TEXT NULL,
                        race_source_concept_id integer NULL REFERENCES CONCEPT (CONCEPT_ID),
                        ethnicity_source_value TEXT NULL,
                        ethnicity_source_concept_id integer NULL REFERENCES CONCEPT (CONCEPT_ID));

When the script runs correctly, you see output like this:

(base) anderson.carl3@home test % ./update-ddl.sh
Adding PRIMARY KEYS to DDL
Adding FOREIGN KEYS to DDL
Creating empty cdm.db database
DONE!
主要语言
HTML
星标
1.1k
派生
515
平均合并
10 天 13 小时
30 天内合并 PR
3

环境准备

这个项目没有提供开发容器、Dockerfile 或贡献指南,环境需要你自己搭建:先看它的 README,通用步骤见我们的新手贡献指南。

从这里开始

  1. 先读完整个 Issue,再读项目的贡献指南。
  2. 在 Issue 下留言说明你要接手 —— 这能避免两个人做同样的事。
  3. Fork 仓库,在一个分支上完成修改。
  4. 提交 Pull Request,并在描述里引用这个 Issue 编号。

OHDSI/CommonDataModel 的其他 Issue

查看 OHDSI/CommonDataModel 的全部 Issue

相似的 Issue

更多 Databases Issue

把新 issue 发到你的邮箱

精选适合新手参与的 GitHub issue 摘要。