Data Formats & Schemas¶

ccpm-scheduler accepts input either as separate CSV files or as a single, unified JSON document. It outputs a CSV schedule, a Markdown build summary, and optional Gantt chart PNG / interactive HTML dependency graph artifacts.

Input Formats¶

1. CSV Files¶

A network can be represented by 2 or 3 CSV files: tasks.csv, resources.csv, and optional calendar.csv.

`tasks.csv`¶

Defines project tasks, duration estimates, dependencies, and resource assignments.

Column	Required	Type	Description
`id`	Yes	String	Unique task ID (e.g. `T1`, `design`).
`name`	No	String	Human-readable name (defaults to `id`).
`realistic_duration`	No*	Integer	Duration estimate including safety (working days; `0` marks a milestone, see below).
`optimal_duration`	No*	Integer	Padding-free aggressive duration estimate (`0` marks a milestone).
`predecessor_ids`	No	String	Semicolon-separated predecessor link string (e.g. `T1`, `T1:SS+2`, `T2:FF`).
`resource_ids`	Yes†	String	Semicolon-separated resource IDs assigned to task (at least 1 required).
`url`	No	String	Optional web URL (passed through to HTML interactive graph and Gantt).

*At least one of realistic_duration or optimal_duration must be specified. If optimal_duration is omitted, it defaults to ceil(realistic_duration / 2).

†Not required for a milestone task (duration 0) — see below.

Legacy column names: predecessors and resources are still accepted as aliases for predecessor_ids and resource_ids (and duration/duration_safe/duration_aggressive for realistic_duration/optimal_duration), but using them raises a W_LEGACY_COLUMNS warning — new files should use the _ids names above. The alias only applies when the current column is absent; if both are present, the current name wins.

Milestone tasks: a task whose duration resolves to exactly 0 is a milestone (a phase gate, a project terminus, an "all done" marker) and is exempt from the "at least 1 resource" rule — it contends for nothing and takes no time. This is the same convention ccpm-scheduler's own internal FINISH node uses; see Network Layout: Guaranteed Single Termination Point for how it's scheduled and merged. A negative or non-numeric duration is still rejected — only exactly 0 qualifies.

Dependency Link Notation: - A or A:FS: Finish-to-Start link (task starts after A finishes). - A:SS+2: Start-to-Start link with 2-day lag (task starts 2 days after A starts). - A:FF: Finish-to-Finish link (task finishes when A finishes). - A:SF-1: Start-to-Finish link with -1 day lag.

`resources.csv`¶

Defines available resources and their default capacity.

Column	Required	Type	Description
`id`	Yes	String	Unique resource ID (e.g. `dev`, `designer`).
`name`	No	String	Resource name (defaults to `id`).
`capacity`	No	Integer	Concurrent capacity limit per day (default: `1`).
`url`	No	String	Optional detail URL.

`calendar.csv` (Optional)¶

Overrides resource capacity over half-open day intervals [from, to).

Column	Required	Type	Description
`resource_id`	Yes	String	Target resource ID.
`from`	Yes	Integer	Start day offset (inclusive).
`to`	Yes	Integer	End day offset (exclusive).
`capacity`	Yes	Integer	Override capacity during interval (`0` = unavailable).

2. Unified JSON Exchange Format¶

The single JSON document format consolidates tasks, resources, calendar, and settings into one structure:

{
  "buffer_method": "cap",
  "resources": [
    { "id": "dev", "name": "Lead Developer", "capacity": 1 },
    { "id": "designer", "name": "UI Designer", "capacity": 1 }
  ],
  "calendar": [
    { "resource_id": "dev", "from": 10, "to": 15, "capacity": 0 }
  ],
  "tasks": [
    {
      "id": "T1",
      "name": "Design Wireframes",
      "realistic_duration": 10,
      "optimal_duration": 5,
      "resources": ["designer"]
    },
    {
      "id": "T2",
      "name": "Implement Frontend",
      "realistic_duration": 14,
      "optimal_duration": 7,
      "predecessors": [
        { "id": "T1", "type": "FS", "lag": 0 }
      ],
      "resources": ["dev"]
    }
  ]
}

JSON network inputs can be passed as a file path or piped via stdin using -:

cat network.json | ccpm-scheduler build - --out-dir plan --json

Output Formats¶

1. `schedule.csv`¶

The generated resource-leveled and buffered schedule.

Column	Description
`id`	Task or buffer ID (e.g. `T1`, `PB`, `FB_T2`). May include the synthetic `FINISH` milestone — see Guaranteed Single Termination Point.
`name`	Row display name.
`type`	Row type: `task` (`FINISH` is a `type: task` row too — duration `0`, no resources), `project_buffer`, or `feeding_buffer`.
`chain`	Chain assignment: `critical`, `feeding-1`, `feeding-2`, or `none`. `FINISH`, when present, is always `critical`.
`start`	Scheduled start day offset.
`finish`	Scheduled finish day offset.
`duration`	Scheduled (optimal) duration in working days.
`realistic_duration`	Original realistic duration estimate (empty for buffers).
`resource_ids`	Semicolon-separated list of assigned resources.
`predecessor_ids`	Semicolon-separated list of predecessor links (includes `X:PB` and `X:FB` buffer attachment links).
`url`	Detail URL passed through from input.

2. `summary.md`¶

A Markdown document generated by build summarizing key project metrics: - Overall project completion promise date (end of Project Buffer). - Critical Chain length and task sequence. - Project Buffer (PB) size and selected sizing method (cap, hchain, or rsem). - Feeding Buffer (FB) breakdown and protected chains. - Calendar availability constraints applied during leveling.

3. Interactive Dependency Network (`project-network.html`)¶

Generated by ccpm-scheduler graph. A single standalone HTML file containing an interactive vis-network view: - Color-coded nodes (Critical Chain = firebrick, Feeding Chains = blue/green/purple, Buffers = gold/khaki). - Resource filtering dropdown menu to isolate tasks assigned to a specific team member. - Detail tooltips showing realistic vs. scheduled duration estimates.

Data Formats & Schemas¶

Input Formats¶

1. CSV Files¶

tasks.csv¶

resources.csv¶

calendar.csv (Optional)¶