github.com/askholme/packer@v0.7.2-0.20140924152349-70d9566a6852/website/source/docs/extend/post-processor.html.markdown (about) 1 --- 2 layout: "docs" 3 page_title: "Custom Post-Processor Development" 4 --- 5 6 # Custom Post-Processor Development 7 8 Post-processors are the components of Packer that transform one artifact 9 into another, for example by compressing files, or uploading them. 10 11 In the compression example, the transformation would be taking an artifact 12 with a set of files, compressing those files, and returning a new 13 artifact with only a single file (the compressed archive). For the 14 upload example, the transformation would be taking an artifact with 15 some set of files, uploading those files, and returning an artifact 16 with a single ID: the URL of the upload. 17 18 Prior to reading this page, it is assumed you have read the page on 19 [plugin development basics](/docs/extend/developing-plugins.html). 20 21 Post-processor plugins implement the `packer.PostProcessor` interface and 22 are served using the `plugin.ServePostProcessor` function. 23 24 <div class="alert alert-block"> 25 <strong>Warning!</strong> This is an advanced topic. If you're new to Packer, 26 we recommend getting a bit more comfortable before you dive into writing 27 plugins. 28 </div> 29 30 31 ## The Interface 32 33 The interface that must be implemented for a post-processor is the 34 `packer.PostProcessor` interface. It is reproduced below for easy reference. 35 The actual interface in the source code contains some basic documentation as well explaining 36 what each method should do. 37 38 39 <pre class="prettyprint"> 40 type PostProcessor interface { 41 Configure(interface{}) error 42 PostProcess(Ui, Artifact) (a Artifact, keep bool, err error) 43 } 44 </pre> 45 46 ### The "Configure" Method 47 48 The `Configure` method for each post-processor is called early in the 49 build process to configure the post-processor. The configuration is passed 50 in as a raw `interface{}`. The configure method is responsible for translating 51 this configuration into an internal structure, validating it, and returning 52 any errors. 53 54 For decoding the `interface{}` into a meaningful structure, the 55 [mapstructure](https://github.com/mitchellh/mapstructure) library is 56 recommended. Mapstructure will take an `interface{}` and decode it into an 57 arbitrarily complex struct. If there are any errors, it generates very 58 human-friendly errors that can be returned directly from the configure 59 method. 60 61 While it is not actively enforced, **no side effects** should occur from 62 running the `Configure` method. Specifically, don't create files, don't 63 create network connections, etc. Configure's purpose is solely to setup 64 internal state and validate the configuration as much as possible. 65 66 `Configure` being run is not an indication that `PostProcess` will ever 67 run. For example, `packer validate` will run `Configure` to verify the 68 configuration validates, but will never actually run the build. 69 70 ### The "PostProcess" Method 71 72 The `PostProcess` method is where the real work goes. PostProcess is 73 responsible for taking one `packer.Artifact` implementation, and transforming 74 it into another. 75 76 When we say "transform," we don't mean actually modifying the existing 77 `packer.Artifact` value itself. We mean taking the contents of the artifact 78 and creating a new artifact from that. For example, if we were creating 79 a "compress" post-processor that is responsible for compressing files, 80 the transformation would be taking the `Files()` from the original artifact, 81 compressing them, and creating a new artifact with a single file: the 82 compressed archive. 83 84 The result signature of this method is `(Artifact, bool, error)`. Each 85 return value is explained below: 86 87 * `Artifact` - The newly created artifact if no errors occurred. 88 * `bool` - If true, the input artifact will forcefully be kept. By default, 89 Packer typically deletes all input artifacts, since the user doesn't generally 90 want intermediary artifacts. However, some post-processors depend on the 91 previous artifact existing. If this is `true`, it forces packer to keep the 92 artifact around. 93 * `error` - Non-nil if there was an error in any way. If this is the case, 94 the other two return values are ignored.