gitplm: A Go repository from Git PLM

Product Lifecycle Management (PLM) in Git.

Additional documents:

GitPLM is a tool and a collection of best practices for managing information needed to manufacture products.

The fundamental thing you want to avoid in any workflow is tedious manual operations that need to made over and over. You want to do something once, and then your tools do it for you from then on. This is the problem that GitPLM solves.

GitPLM does several things:

combines source BOMs with the partmaster to generate BOMs with manufacturing information.
automate the generation of release/manufacturing information
create combined BOMs that include parts from all sub-assemblies
gathers release data for all custom components in the design into one directory for release to manufacturing.

An example output is shown below:

GitPLM is designed for small teams building products. We leverage Git to track changes and use simple file formats like CSV to store BOMs, partmaster, etc.

Video overview

GitPLM overview

Installation

You can download a release for your favorite platform. This tool is a self-contained binary with no dependencies.

Alternatively, you can:

go intstall github.com/git-plm/gitplm@latest

clone the Git repo and run: go run .

Usage

Type gitplm from a shell to see commandline options:

Usage of gitplm:
  -release string
	      Process release for IPN (ex: PCB-056-0005, ASY-002-0023)
  -version int
        display version of this application

Part Numbers

Each part used to make a product is defined by a IPN (Internal Part Number). The convention used by GitPLM is: CCC-NNN-VVVV

CCC: major category (RES, CAP, DIO, etc)
NNN: incrementing sequential number for each part
VVVV: variation to code variations of a parts typically with the same datasheet (resistance, capacitance, regulator voltage, IC package, etc.) Also used to encode the version of custom parts or assemblies.

Partmaster

A single partmaster.csv file is used for the entire organization and contains internal part numbers (IPN) for all assets used to build a product. For externally sourced parts, purchasing information such as manufacturer part number (MPN) is also included.

If multiple sources are available for a part, these can be entered on additional lines with the same IPN, and different Manufacturer/MPN specified. GitPLM will merge other fields like Description, Value, etc so these only need to be specified on one of the lines. The Priority column is used to select the preferred part (lowest number wins). If no Priority is set, it defaults to 0 (highest priority). Currently, GitPLM picks the highest priority part and populates that in the output BOM. In the future, we could add additional columns for multiple sources.

CAD tool libraries should contain IPNs, not MPNs. Why not just put MPNs in the CAD database? The fundamental reason is that a single part may be used in 100's of different places and dozens of assemblies. If you need to change a supplier for a part, you don't want to manually modify a dozen designs, generate new BOMs, etc. This is manual, tedious, and error prone. What you want to do is change the manufacturer information in the partmaster and then automatically generate new BOMs for all affected products. Because the BOMs are stored in Git, it is easy to review what changed.

Components you manufacture

A product is typically a collection of custom parts you manufacture and off-the-shelf parts you purchase. Custom parts are identified by the following CCCs:

Code	Description
PCA	Printed Circuit Assembly. The version is incremented any time the BOM for the assembly changes.
PCB	Printed Circuit board. This category identifies the bare PCB board.
ASY	Assembly (can be mechanical or top level subassembly -- typically represented by BOM and documentation). Again, the variation is incremented any time a BOM line item changes. You can also use product specific prefixes such as GTW (gateway).
DOC	standalone documents
DFW	data -- firmware to be loaded on MCUs, etc
DSW	data -- software (images for embedded Linux systems, applications, programming utilities, etc)
DCL	data -- calibration data for a design
FIX	manufacturing fixtures

If IPN with the above category codes are found in a BOM, GitPLM looks for release directory that matches the IPN and then soft-links from the release directory to the sub component release directory. In this way we build up a hierarchy of release directories for the entire product.

Source and Release directories

For parts you produce, GitPLM scans the directory tree looking for source directories which are identified by one or both of the following files:

an input BOM. Ex: ASY-023.csv
a release configuration file. Ex: PCB-019.yml

If either of these is found, GitPLM considers this a source directory and will use this directory to generate release directories.

A source directory might contain:

A PCB designs
Application source code
Firmware
Mechanical design files
Test procedures
User documentation
Test Fixures/Procedures

Release directories are identified by a full IPN. Examples:

PCA-019-0012
ASY-012-0002
DOC-055-0006

Special Files

The following files will be copied into the release directory if found in the project directory:

MFG.md: contains notes for manufacturing
CHANGELOG.md: contains a list of changes for each version. See keep a changelog for ideas on how to structure this file. Every source directory should have a CHANGELOG.md.

Release configuration

A release configuration file (CCC-NNN.yml) in the source directory can be used to customize the release process.

The file format is YAML, and an example is shown below:

remove:
  - cmpName: Test point
  - cmpName: Test point 2
  - ref: D12
add:
  - cmpName: "screw #4,2"
    ref: S3
    ipn: SCR-002-0002
hooks:
  - date -Iseconds > {{ .RelDir }}/timestamp.txt
  - |
    echo "processing {{ .SrcDir }}"
    echo "hi #1"
    echo "hi #2"
copy:
  - gerber
  - mfg
  - pcb.schematic
required:
  - PCA-019-0002_ibom.html

The following template variables are available:

RelDir: the release directory that GitPLM is generating
SrcDir: the source directory GitPLM is pulling information from

Supported operations:

remove: remove a part from a BOM
add: add a part to a BOM
copy: copy a file or dir to the release directory
hooks: run shell scripts (currently Linux/MacOS only). Can be used to build software, generate PDFs, etc.
required: looks for required files in the release directory and stops with an error if they are not found. This is used to check that manually generated files have been populated.

The release process should be automated as much as possible to process the source files and generate the release information with no manual steps.

Examples

See the examples folder. You can run commands like to exercise GitPLM:

go run . -release ASY-001-0000
go run . -release PCB-019-0001

go run . is used when working in the source directory. You can replace this with gitplm if you have it installed.

Principles

manual operations/tweaks to machine generated files are bad. If changes are made (example a BOM line item add/removed/changed), this needs to be defined declaratively and then this change applied by a program. Ideally this mechanism is also idempotent, so we describe where we want to end up, not steps to get there. The program can determine how to get there.
the number of parts used in a product is bounded, and can easily fit in computer memory (IE, we probably don't need a database for small/mid sized companies)
the total number of parts a company may use (partmaster) is also bounded, and will likely fit in memory for most small/mid sized companies.
tracking changes is important
review is important, thus Git workflow is beneficial
ASCII (text) files are preferred as they can be manually edited and changes easily review in Git workflows.
versions are cheap -- VVVV should be incremented liberally.
PLM software should not be tied to any one CAD tool, but should be flexible enough to work with any CAD output.

Additional notes

use CSV files for partmaster and all BOMs.
- rational: can be read and written by excel, libreoffice, or by machine
- rational: easy to get started
versions in part numbers are sequential numbers: (0, 1, 2, 3, 4)
- rational: easy to use in programs, sorting, etc
CAD BOMs are never manually "scrubbed". If additional parts are needed in the assembly, create a higher level BOM that includes the CAD generated BOM, or create a *.yml file to declaratively describe modifications to the BOM.
- rational: since the CAD program generates the BOM in the first place, any manual processing of this BOM will only lead to mistakes.
CSV files should be delimited with ';' instead of ','.
- rational: comma is useful in lists, descriptions, etc.
Tooling is written in Go.
- rational:
  - Go programs are reasonably reliable
  - it is easy to generate standalone binaries for most platforms with no dependencies
  - Go is fast
  - Go is easy to read and learn.
  - The Go package ecosystem is quite extensive. go-git may be useful for tight integration with Git.
  - Program can be started as a command line program, but eventually grow into a full-blown web application.

git-plm/gitplm