appendix-config.qmd

---
title: "C: Configuration YAML"
---



The below is an example of configuration options configured in more human readable YAML format.

```yaml

# ETL Pipeline Configuration

general:
  hostkeys: 
    - INB112
    # - N420
  folder_name: '' # default to hostkey if empty

file_paths:
  root_dir: '.'  # default to current working directory
  nodes_folder_url: # (Optional) override for dynamic lookup) eg "https://drive.google.com/drive/folders/1Rc3vQCF6CwxV3yNjfUTWXv61BgYD1j_3"
  relationships_folder_url: # (Optional) override for dynamic lookup) eg."https://drive.google.com/drive/folders/1w_ea6ETzRcdYz71crLxL9khjLrEfcbuH"
  gdrive_root_folder_url: "1iWkeTubJ0xZ6I728emoj9BkqZm7dL2fq"
  gdrive_folder_name: # Leave commented out to use default (hostkey)
  google_credentials_path: 'credentials/graph-diss-dbbdbb5e5d00.json'
  department_source: 'node-dept-all.csv'
  archibus_source: 'archibus.csv'

data_processing:
  chunk_size: 20000
  temp_tables_sql_file: "create_temp_tables.sql"
  node_output_filename_template: "node-{node}-processed.csv"
  rel_output_filename_template: "rel-{relationship}-processed.csv"

neo4j:
  #max_connection_retries: 5
  #max_transaction_retry_time: 30
  schema:
    apply: True
    type: 'dynamic' # Options: 'dynamic', 'custom'
    custom_path: ''
  batch_size: 1000

logging:
  log_level: "INFO" # Options: DEBUG, INFO, WARNING, ERROR, CRITICAL 

nodes:
  department: 
    filename_pattern: "node-dept-all*.csv"
    dept_join_col: null 
    node_suffix: 'dept'
    node_id: "deptSplusID"
  module: 
    filename_pattern: "node-module-by-pos-temp*.csv"
    dept_join_col: "modSplusDeptID"
    node_suffix: "mod"
    node_id: "modSplusID"             
  room: 
    filename_pattern: "node-room-by-pos-temp*.csv"
    dept_join_col: null
    node_suffix: 'room'
    node_id: "roomSplusID"   
  programme: 
    filename_pattern: "node-pos-by-pos-temp*.csv"
    dept_join_col: "posSplusDeptID"
    node_suffix: "pos"
    node_id: "posSplusID"
  activityType: 
    filename_pattern: "node-activitytype-by-pos-temp*.csv"
    dept_join_col: 'actTypeDeptSplusID'
    node_suffix: 'actType'
    node_id: 'actTypeSplusID'
  staff: 
    filename_pattern: "node-staff-by-pos-temp*.csv"
    dept_join_col: "staffDeptSplusID"
    node_suffix: "staff"
    dtype:
      staffSplusID: str
      staffID: str 
    node_id: "staffSplusID"
  student: 
    filename_pattern: "node-student-by-pos-temp*.csv"
    dept_join_col: "stuDeptSplusID"
    node_suffix: "stu"
    dtype: 
      stuSplusID: str
      studentID: str
    node_id: "stuSplusID"
  activity: 
    filename_pattern: "node-activity-by-pos-temp*.csv"
    dept_join_col: null
    node_suffix: null
    dtype:
      actSplusID: str
      actTypeSplusID: str 
      actRoomSplusID: str
      actStaffSplusID: str 
      actStuSplusID: str 
      actStartDateTime: str
      actEndDateTime: str
      actFirstActivityDate: str
      actLastActivityDate: str
      actWhenScheduled: str
    node_id: "actGraphID" 

relationships:
  activity_module: 
    filename_pattern: "rel-activity-module-by-pos-temp*.csv"
    node1_col: "actSplusID"
    node2_col: "modSplusID"
    relationship: "BELONGS_TO"
  activity_room: 
    filename_pattern: "rel-activity-room-by-pos-temp*.csv"
    node1_col: "actSplusID"
    node2_col: "roomSplusID"
    relationship: "OCCUPIES"
  activity_staff: 
    filename_pattern: "rel-activity-staff-by-pos-temp*.csv"
    node1_col: "staffSplusID"
    node2_col: "actSplusID"
    relationship: "TEACHES"
  activity_student: 
    filename_pattern: "rel-activity-student-by-pos-temp*.csv"
    node1_col: "stuSplusID"
    node2_col: "actSplusID"
    relationship: "ATTENDS"
  activity_activityType: 
    filename_pattern: "relActivityActType*.csv"
    node1_col: "actSplusID"
    node2_col: "actActivityTypeSplusID"
    relationship: "HAS_TYPE"
  module_programme: 
    filename_pattern: "rel-mod-pos-by-pos-temp*.csv"
    node1_col: "modSplusID"
    node2_col: "posSplusID"
    relationship: "BELONGS_TO"
    properties: 
      - "modType"

data_type_mapping:
  activity:
    actStartDateTime: ['datetime', '%Y-%m-%d %H:%M:%S']
    actEndDateTime: ['datetime', '%Y-%m-%d %H:%M:%S']
    actFirstActivityDate: ['date2', '%Y-%m-%d']
    actLastActivityDate: ['date2', '%Y-%m-%d']
    actPlannedSize: 'int'
    actRealSize: 'int'
    actDuration: 'int'
    actDurationInMinutes: 'int'
    actNumberOfOccurrences: 'int'
    actWhenScheduled: ['datetime', '%Y-%m-%d %H:%M:%S']
    actStartDate: ['date', '%Y-%m-%d']
    actEndDate: ['date', '%Y-%m-%d']
    actStartTime: 'time'
    actEndTime: 'time'
    actScheduledDay: 'int'
  room:
    roomCapacity: 'int'

display_name_mapping:
  activity: "actName"
```