This would add a command that would print out the specific settings that would be applied to a target for a specific config. This would combine

• target build settings
• target xcconfig file
• project build settings
• project xcconfig file

To begin with this wouldn't include settings that are applied by the top most level (the SDK) as we don't have easy access to that.

This lets the Carthage folder reside in other places than the project root

This is by specifying multiple platforms for a target in the spec, which would generate multiple targets, each with a different platform

Add the ability to add a TestTargetID in project attributes
You can already set the TEST_TARGET_NAME build setting but that doesn't change the "Host application" in the Xcode UI

I often use auto-increment build number script and some scripts that should invoked only when archiving.

Examples:
http://crunchybagel.com/auto-incrementing-build-numbers-in-xcode/
https://gist.github.com/sekati/3172554
https://github.com/realm/realm-cocoa/blob/master/scripts/strip-frameworks.sh

Those kind of scripts should run only when archiving (well, just personally I guess so). So I believe 'Run script only when installing' option would be nice...

Generate pre and post build run script build phases

It would be nice if there were a way to clear the default settings XcodeGen provides. This is useful if you define all your settings in xcconfig files, so when you generate the project, you don't want its default settings to shadow yours. I almost expected it to with this configuration:

name: Project
settings:
  base: []
targets:
  Project:
    type: application
    platform: iOS

But I understand these base settings are just additive to the project. In the meantime I might use xcodeproj to wipe these settings out after project generation.

Should be able to generate a spec from an existing project for easy migration

If a storyboard localized for some language, it should be grouped into one variant group.

XcodeGen generates references same level.

expectation

VG1473702401 /* Main.storyboard */ = {
	isa = PBXVariantGroup;
	children = (
		FR3676338402 /* Base */,
	);
	name = Main.storyboard;
	sourceTree = "<group>";
};
VG2365450901 /* Main.strings */ = {
	isa = PBXVariantGroup;
	children = (
		FR7070045401 /* en */,
	);
	name = Main.strings;
	sourceTree = "<group>";
};

should be

VG1473702401 /* Main.storyboard */ = {
	isa = PBXVariantGroup;
	children = (
		FR3676338402 /* Base */,
		FR7070045401 /* en */,
	);
	name = Main.storyboard;
	sourceTree = "<group>";
};

I have Xcode 9 GM installed, and have ensured that I am using them from the command-line. When I do the following homebrew commands:

        brew update
        brew tap yonaskolb/XcodeGen https://github.com/yonaskolb/XcodeGen.git
        brew install XcodeGen

then I am getting the following errors both on my local laptop and our CI servers:

Updated 2 taps (caskroom/cask, homebrew/core).
==> New Formulae
azure-cli@1
json-table
==> Updated Formulae
akamai
angular-cli
apache-opennlp
axel
azure-cli
bandcamp-dl
...
webpack
wpscan
xorriso
xtensor
you-get
youtube-dl
==> Deleted Formulae
kes
==> Tapping yonaskolb/xcodegen
Cloning into '/usr/local/Homebrew/Library/Taps/yonaskolb/homebrew-xcodegen'...
Tapped 1 formula (99 files, 2.7MB)
==> Installing xcodegen from yonaskolb/xcodegen
==> Downloading https://github.com/yonaskolb/XcodeGen/archive/0.6.1.tar.gz
==> Downloading from https://codeload.github.com/yonaskolb/XcodeGen/tar.gz/0.6.1
==> Building XcodeGen
==> swift build -c release -Xlinker -rpath -Xlinker @executable_path -Xswiftc -static-stdlib
Last 15 lines from /Users/builduser/Library/Logs/Homebrew/xcodegen/01.swift:
2017-09-18 18:59:50 +0000

swift build -c release -Xlinker -rpath -Xlinker @executable_path -Xswiftc -static-stdlib

error: manifest parse error(s):
sandbox-exec: sandbox_apply_container: Operation not permitted

If reporting this issue please do so to (not Homebrew/brew or Homebrew/core):
yonaskolb/xcodegen

/usr/local/Homebrew/Library/Homebrew/utils/github.rb:226:in `raise_api_error': curl failed! curl: (22) The requested URL returned error: 422 Unprocessable Entity (GitHub::Error)
curl: (3) <url> malformed
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:184:in `open'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:284:in `search'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:231:in `search_issues'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:244:in `issues_for_formula'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:369:in `fetch_issues'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:365:in `issues'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:419:in `dump'
	from /usr/local/Homebrew/Library/Homebrew/brew.rb:133:in `rescue in <main>'
	from /usr/local/Homebrew/Library/Homebrew/brew.rb:26:in `<main>'

allow easy carthage dependencies

dependencies:
    carthage: Result

Currently if I generate a project that has a xcconfig set for the configuration, and also includes the directory the xcconfig file lives in in the sources of a target, the project ends up in a broken state:

This is the configuration for this case:

name: Project
configs:
  Debug: debug
configFiles:
  Debug: xcconfigs/debug.xcconfig
targets:
  Project:
    type: application
    platform: iOS
    sources: xcconfigs

Sample project (with the generated project) as an example: XcodeGenTest.zip

This is for sharing build settings in multiple places

This would allow to linking to frameworks at a certain path

The readme should contain a proper documentation. 👍

Using XcodeGen 0.3.0 and Xcode 8.3.3.

If one of files name contains "@", xcodegen generates non-parsable project by Xcode, somehow.
So it cannot deal with retina filenames such as [email protected], [email protected].
Adding those kind of assets to *.xcassets wouldn't really solve this problem as I intend not to use it, since using xcassets forces UIImage to be cached.

example:

name: FooBar
targets:
  - name: FooBar
    type: application
    platform: iOS
    sources: 
      - Assets

If "Assets" directory or its subdirectory contains a file that has "@" in its name, Xcode cannot parse generated project.
The other tool, struct, can correctly generates a project in the same situation, just fwiw.

lgtm $ xcodegen-debug
📋  Loaded spec:
  Name: LGTM
  Targets:
    🖥  application: LGTM
    📱  framework: LGTMKit_iOS
    🖥  framework: LGTMKit_macOS
    📺  framework: LGTMKit_tvOS
    ⌚️  framework: LGTMKit_watchOS
    📱  application: LGTM_iOS
2 Spec validations errors:
	- Target "LGTM" has invalid dependency: "LGTMKit"
	- Target "LGTM_iOS" has invalid dependency: "LGTMKit"

How could I generate an empty(buildSettings = { };) framework target?
I don't want each target for every platform.

Although embedding frameworks is working flawlessly, it seems Code Sign on Copy will be always false (checked off) even in case specifying it manually by: codeSign: true.

If multiple targets have share the same source folder

It seems that https://github.com/swift-xcode/xcodeproj supports PBXAggregateTarget (?), but XcodeGen doesn't.
Priority would be low though, I often use the kind of target in framework projects to invoke archive script.

An optional file watcher that automatically regenerates the project when file system changes. Should be able to specify watched files with a glob

Currently, if you have more than one configs xcodegen would make schemes Shared.
If there is an option to make a scheme Shared checked on, it would be great.

Ability to support spec includes. This would allow you to seperate the spec into multiple files, to for example seperate all the setting presets for sharing amongst multiple specs

This would allow to linking of system SDK frameworks

This will make the Carthage copy-frameworks script run only if the output files are missing. As suggested in the Carthage readme

This results in projects that don't have default build settings applied for config type, platform or product type

With this config:

name: Project
targets:
  Project:
    type: application
    platform: iOS
    sources:
      - Sources/foo.swift

And this file structure:

.
├── Sources
│   └── Foo.swift
└── project.yml

1 directory, 2 files

xcodegen fails with this error message:

📋  Loaded spec:
  Name: Project
  Targets:
    📱  application: Project
Generation failed: The file “Foo.swift” couldn’t be opened.

Should you be able to reference a path like this absolutely? This is useful for when you have more complex file structures where sources and tests are nearby, meaning you can't reference the entire directory or you'll include test files in the non-test target.

Sample project: XcodeGenTest2.zip

Generation fails unless App.xcodeproj/project.xcworkspace/contents.xcworkspacedata is present.

workaround:

$ mkdir LGTM.xcodeproj/project.xcworkspace
$ :> LGTM.xcodeproj/project.xcworkspace/contents.xcworkspacedata
$ xcodegen
📋  Loaded spec:
  Name: LGTM
  Targets:
    🖥  application: LGTM
    📱  framework: LGTMKit_iOS
    🖥  framework: LGTMKit_macOS
    📱  application: LGTM_iOS
⚙️  Generated project
💾  Saved project to LGTM.xcodeproj

Motivation

Auto apply xcconfig configurations like this.

Current Problems

I notice documentations do mention about xcconfig but looks like it's not working as expected.

• Setting configFiles does not affect anything. (Bug? or maybe my spec is wrong)
• There is no way to set base xcconfigs for PBXProject.
• xcconfigs not being added to project

Proposal

So this is my proposal.

• Support baseConfiguration for PBXProj
• Rename target's configFiles with baseConfigurations for consistency.
• (Optional) Rename configs to configurations since it looks like xcconfig and is misleading.

name: MyProject
configurations:
  Debug:
    baseConfiguration: configs/Debug.xcconfig
  OTA:
    baseConfiguration: configs/OTA.xcconfig
  Release:
    baseConfiguration: configs/Release.xcconfig
targets:
  MyApp:
    type: application
    platform: iOS
    sources: Sources/MyApp
    settings:
      PRODUCT_BUNDLE_IDENTIFIER: jp.toshi0383.hello
    baseConfigurations:
      Debug: configs/MyApp-Debug.xcconfig
      OTA: configs/MyApp-OTA.xcconfig
      Release: configs/MyApp-Release.xcconfig

If you guys feel happy with this outline, I can start working on a PR. 💪

Adding specific directories as Folder References would be really nice feature.
I'm referring to the thing below:

This is just a question about include feature.
How is it different from simply using xcconfig's #include and Xcode's $(inherited) ?

Using the instructions from the README I currently get this:

brew install xcodegen
==> Installing xcodegen from yonaskolb/xcodegen
==> Downloading https://github.com/yonaskolb/XcodeGen/archive/0.6.1.tar.gz
Already downloaded: /Users/dbeard/Library/Caches/Homebrew/xcodegen-0.6.1.tar.gz
==> Building XcodeGen
==> swift build -c release -Xlinker -rpath -Xlinker @executable_path -Xswiftc -static-stdlib
Last 15 lines from /Users/dbeard/Library/Logs/Homebrew/xcodegen/01.swift:
2017-09-13 12:02:33 -0700

swift build -c release -Xlinker -rpath -Xlinker @executable_path -Xswiftc -static-stdlib

error: manifest parse error(s):
sandbox-exec: sandbox_apply_container: Operation not permitted

If reporting this issue please do so to (not Homebrew/brew or Homebrew/core):
yonaskolb/xcodegen

/usr/local/Homebrew/Library/Homebrew/utils/github.rb:226:in `raise_api_error': curl failed!  (GitHub::Error)
curl: (22) The requested URL returned error: 422 Unprocessable Entity
curl: (3) <url> malformed
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:184:in `open'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:284:in `search'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:231:in `search_issues'
	from /usr/local/Homebrew/Library/Homebrew/utils/github.rb:244:in `issues_for_formula'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:369:in `fetch_issues'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:365:in `issues'
	from /usr/local/Homebrew/Library/Homebrew/exceptions.rb:419:in `dump'
	from /usr/local/Homebrew/Library/Homebrew/brew.rb:133:in `rescue in <main>'
	from /usr/local/Homebrew/Library/Homebrew/brew.rb:26:in `<main>'

At the moment targets is list. That was done to preserve the order of generated targets in the xcode project.
There would be some upsides to making it a map of names to targets however:

• the include feature could edit and merge targets by name
• more natural yaml styl
• enforce the target name
• better visual seperation

The points are very minor other than the first one which is a real consideration.
Is having the order of targets in the project worth it? Thoughts anyone?

Here is a visual difference

list - Array

targets:
  - name: MyTarget
    platform: iOS
    type: application
  - name: MyFramework
    platform: iOS
    type: framework

map - Dictionary

targets:
  MyTarget:
    platform: iOS
    type: application
  MyFramework:
    platform: iOS
    type: framework 

In either case targets in the ProjectSpec struct would still be an array, this just affects yaml

Is there any difference from using xcconfig?
What are pros/cons ?

Related: #64 (comment)

Motivation

With spec like this,

targets:
  LGTM:
    platform: macOS
    type: application
    sources: LGTM
    dependencies:
      - target: LGTMKit_macOS
  LGTMKit:
    type: framework
    platform: [iOS, macOS]
    sources: LGTMKit
    dependencies:
      - carthage: APIKit

Generate xcodeproj like this.

Related: #82

This would lint the project after generation with https://github.com/carambalabs/xcodeprojlint to make sure no issues exist

I find it useful to include priority, difficulty and type useful to organize issues. It's also good for welcoming contributors in the project (e.g. they can start with the simple issues to get some context on the project).

This is the setup that I've used for xcodeproj, what do you think @yonaskolb?

Installation can be simplified by using Homebrew. The purpose of this issue is to provide a Homebrew formula that tells Homebrew how to install XcodeGen

Hello,
Thanks for your this great solution that I really wanted.

Today I tried to generate my project using XcodeGen.
I faced the problem that Xcode shows the message This app was unable to be installed after build the project for the simulator.

I reproduced it on TestProject in this repository.

1. xcodegen --spec spec.yml
2. Open GeneratedProject.xcodeproj
3. Select the simulator like iPhone 8 Plus
4. Run

It can be fixed by removing frameworks from Copy Files in Build Phases or remove the line https://github.com/yonaskolb/XcodeGen/blob/master/Sources/XcodeGenKit/PBXProjGenerator.swift#L236

I don't know this is correct way to fix it. But do frameworks need to be copied when these add as Embedded Framework?

I just had an idea. What do you think about being able to generate simple projects from the command line with XcodeGen. Something like:

xcodegen init -platform iOS -name MyTest --no-tests

That would create the xcodegen.yml file and generate the project from it. How does that sound? @yonaskolb

I often use Swift dynamic framework and specify them in xcodegen.yml like below:

name: Foo
targets:
  - name: Bar
    type: application
    platform: iOS
    sources: 
      - Sources
    dependencies:
      - framework: Frameworks/Foo.framework
      - framework: Frameworks/Bar.framework

XcodeGen correctly handle this and add those frameworks to Linked Frameworks and Libraries, but it seems we do not have an option to have them as Embedded Binaries?

I am working on a a Swift project and I need to be able to generate a Playground inserting my own file into it to get some autocomplete

example xcodegen.yml:

name: FooBar
targets:
  - name: FooBar
    type: application
    platform: iOS
    sources: 
      - Sources
    postbuildScripts:
      - path: Scripts/strip-frameworks.sh
        name: Strip Unused Architectures from Frameworks
        runOnlyWhenInstalling: true

Xcode now says Project *.xcodeproj cannot be opened because the project file cannot be parsed.. Script used above is: https://raw.githubusercontent.com/realm/realm-cocoa/master/scripts/strip-frameworks.sh
Although in the example I'm specifying path of the script, writing the script as string in yamls causes the same problem naturally.

Now, here is the concerned excerpt from generated project:

/* Begin PBXShellScriptBuildPhase section */
        SSBP35707201 /* Run Script */ = {
            isa = PBXShellScriptBuildPhase;
            buildActionMask = 2147483647;
            files = (
            );
            inputPaths = (
            );
            name = "Strip Unused Architectures from Frameworks";
            outputPaths = (
            );
            runOnlyForDeploymentPostprocessing = 1;
            shellPath = /bin/sh;
            shellScript = "################################################################################\n#\n# Copyright 2015 Realm Inc.\n#\n# Licensed under the Apache License, Version 2.0 (the \"License\");\n# you may not use this file except in compliance with the License.\n# You may obtain a copy of the License at\n#\n# http://www.apache.org/licenses/LICENSE-2.0\n#\n# Unless required by applicable law or agreed to in writing, software\n# distributed under the License is distributed on an \"AS IS\" BASIS,\n# WITHOUT WARRANTIES OR CONDITIONS OF ANY KIND, either express or implied.\n# See the License for the specific language governing permissions and\n# limitations under the License.\n#\n################################################################################\n\n# This script strips all non-valid architectures from dynamic libraries in\n# the application's `Frameworks` directory.\n#\n# The following environment variables are required:\n#\n# BUILT_PRODUCTS_DIR\n# FRAMEWORKS_FOLDER_PATH\n# VALID_ARCHS\n# EXPANDED_CODE_SIGN_IDENTITY\n\n\n# Signs a framework with the provided identity\ncode_sign() {\n  # Use the current code_sign_identitiy\n  echo \"Code Signing $1 with Identity ${EXPANDED_CODE_SIGN_IDENTITY_NAME}\"\n  echo \"/usr/bin/codesign --force --sign ${EXPANDED_CODE_SIGN_IDENTITY} --preserve-metadata=identifier,entitlements $1\"\n  /usr/bin/codesign --force --sign ${EXPANDED_CODE_SIGN_IDENTITY} --preserve-metadata=identifier,entitlements \"$1\"\n}\n\n# Set working directory to product’s embedded frameworks \ncd \"${BUILT_PRODUCTS_DIR}/${FRAMEWORKS_FOLDER_PATH}\"\n\nif [ \"$ACTION\" = \"install\" ]; then\n  echo \"Copy .bcsymbolmap files to .xcarchive\"\n  find . -name '*.bcsymbolmap' -type f -exec mv {} \"${CONFIGURATION_BUILD_DIR}\" \;\nelse\n  # Delete *.bcsymbolmap files from framework bundle unless archiving\n  find . -name '*.bcsymbolmap' -type f -exec rm -rf \"{}\" +\;\nfi\n\necho \"Stripping frameworks\"\n\nfor file in $(find . -type f -perm +111); do\n  # Skip non-dynamic libraries\n  if ! [[ \"$(file \"$file\")\" == *\"dynamically linked shared library\"* ]]; then\n    continue\n  fi\n  # Get architectures for current file\n  archs=\"$(lipo -info \"${file}\" | rev | cut -d ':' -f1 | rev)\"\n  stripped=\""\n  for arch in $archs; do\n    if ! [[ \"${VALID_ARCHS}\" == *\"$arch\"* ]]; then\n      # Strip non-valid architectures in-place\n      lipo -remove \"$arch\" -output \"$file\" \"$file\" || exit 1\n      stripped=\"$stripped $arch\"\n    fi\n  done\n  if [[ \"$stripped\" != \"" ]]; then\n    echo \"Stripped $file of architectures:$stripped\"\n    if [ \"${CODE_SIGNING_REQUIRED}\" == \"YES\" ]; then\n      code_sign \"${file}\"\n    fi\n  fi\ndone\n";
        };
/* End PBXShellScriptBuildPhase section */

I have these Localizable settings under resources/language directory.

resources/language/
├── Base.lproj
│   └── Localizable.strings
└── ja.lproj
    └── Localizable.strings

I want to import them into separated target.

targets:
  LGTM:
    platform: macOS
    type: application
    sources:
      - LGTM
      - resources/language
  LGTM_iOS:
    platform: iOS
    type: application
    sources:
      - LGTM_iOS
      - resources/language

This spec ends-up with the files only added to LGTM target.

Also on Xcode, both files are displayed as same name, so maybe it's better if we could display the parent directory names for Localizable.strings.