draft create for Maven projects gets confused by the mvnw and mvnw.cmd files: "Draft detected the primary language as Shell..."

[cyrille-leclerc]

Problem Description

Java Maven projects using Takari's mvnw wrapper don't get properly discovered by draft create. They are seen as "Shell" applications.

Takari's mvnw wrapper is a very popular extension to Maven, it is included by default by Spring initializer (http://start.spring.io/) and by other leading Java project

$ draft create
--> Draft detected the primary language as Shell with 73.597919% certainty.
Error: Could not find a pack Q_Q

To reproduce

git clone https://github.com/Azure/draft.git
cd draft/examples/example-java
wget https://raw.githubusercontent.com/takari/maven-wrapper/maven-wrapper-0.3.0/mvnw
draft create

Error message

--> Draft detected the primary language as Shell with 73.597919% certainty.
Error: Could not find a pack Q_Q

Desired Message

--> Draft detected the primary language as Java with 92.205567% certainty.
--> Ready to sail

Workaround

Before running draft create, temporarily move mvnw and mvnw.cmdoutside of the project. Once draft create is successful, move mvnw and mvnw.cmd back in the project.

[cyrille-leclerc]

Seem to be related to the file pkg/linguist/static.go that comes from https://github.com/github/linguist/blob/v6.0.1/lib/linguist/languages.yml

I'll try to declare mvnw next to the already declared gradlew.

https://github.com/github/linguist/blob/v6.0.1/lib/linguist/languages.yml#L4274

[bacongobbler]

Given that mvnw is primarily written in bash, one could argue that linguist is "working as intended" in this case since it has not been trained against mvnw. In the meantime, you can override certain files using the .gitattributes file as per the troubleshooting documentation.

Specifically, this should work for your example:

$ cat .gitattributes
mvnw linguist-vendored

That way, mvnw is ignored as a vendored file. Moving to the future, you may want to look into adding mvnw as a vendored file at https://github.com/github/linguist/blob/master/lib/linguist/vendor.yml. That way users won't need to add this file to their .gitattributes file every time they wish to incorporate mvnw into their project.

Does that help clarify things?

[cyrille-leclerc]

Thanks @bacongobbler . I'll try to get "mvnw" added to https://github.com/github/linguist/blob/master/lib/linguist/vendor.yml

[cyrille-leclerc]

FYI I have just submitted github-linguist/linguist#4042

[bacongobbler]

Sounds good! Now that merge has made its way into linguist, it shouldn't be too hard to re-train our bayesian classifier based on that.