How to define a variable in a Dockerfile?

Question

In my Dockerfile, I would like to define variables that I can use later in the Dockerfile.

I am aware of the ENV instruction, but I do no want these variables to be environment variables.

Is there a way to declare variables at Dockerfile scope?

Answer 1

You can use ARG - see https://docs.docker.com/engine/reference/builder/#arg

The ARG instruction defines a variable that users can pass at build-time to the builder with the docker build command using the --build-arg <varname>=<value> flag. If a user specifies a build argument that was not defined in the Dockerfile, the build outputs an error.

Can be useful with COPY during build time (e.g. copying tag specific content like specific folders) For example:

ARG MODEL_TO_COPY
COPY application ./application
COPY $MODEL_TO_COPY ./application/$MODEL_TO_COPY

While building the container:

docker build --build-arg MODEL_TO_COPY=model_name -t <container>:<model_name specific tag> .

Answer 2

To answer your question:

In my Dockerfile, I would like to define variables that I can use later in the Dockerfile.

You can define a variable with:

ARG myvalue=3

Spaces around the equal character are not allowed.

And use it later with:

RUN echo $myvalue > /test

Answer 3

To my knowledge, only ENV allows that, as mentioned in "Environment replacement"

Environment variables (declared with the ENV statement) can also be used in certain instructions as variables to be interpreted by the Dockerfile.

They have to be environment variables in order to be redeclared in each new containers created for each line of the Dockerfile by docker build.

In other words, those variables aren't interpreted directly in a Dockerfile, but in a container created for a Dockerfile line, hence the use of environment variable.

---

This day, I use both ARG (docker 1.10+, and docker build --build-arg var=value) and ENV.
Using ARG alone means your variable is visible at build time, not at runtime.

My Dockerfile usually has:

ARG var
ENV var=${var}

In your case, ARG is enough: I use it typically for setting http_proxy variable, that docker build needs for accessing internet at build time.

---

Christopher King adds in the comments:

Watch out!
The ARG variable is only in scope for the "stage that it is used" and needs to be redeclared for each stage.

He points out to Dockerfile / scope

An ARG variable definition comes into effect from the line on which it is defined in the Dockerfile not from the argument’s use on the command-line or elsewhere.

For example, consider this Dockerfile:

FROM busybox
USER ${user:-some_user}
ARG user
USER $user
# ...

A user builds this file by calling:

docker build --build-arg user=what_user .

The USER at line 2 evaluates to some_user as the user variable is defined on the subsequent line 3.
The USER at line 4 evaluates to what_user as user is defined and the what_user value was passed on the command line.
Prior to its definition by an ARG instruction, any use of a variable results in an empty string.

An ARG instruction goes out of scope at the end of the build stage where it was defined.
To use an arg in multiple stages, each stage must include the ARG instruction.

Answer 4

If the variable is re-used within the same RUN instruction, one could simply set a shell variable. I really like how they approached this with the official Ruby Dockerfile.

Answer 5

You can use ARG variable defaultValue and during the run command you can even update this value using --build-arg variable=value. To use these variables in the docker file you can refer them as $variable in run command.

Note: These variables would be available for Linux commands like RUN echo $variable and they wouldn't persist in the image.

Answer 6

Late to the party, but if you don't want to expose environment variables, I guess it's easier to do something like this:

RUN echo 1 > /tmp/__var_1
RUN echo `cat /tmp/__var_1`
RUN rm -f /tmp/__var_1

I ended up doing it because we host private npm packages in aws codeartifact:

RUN aws codeartifact get-authorization-token --output text > /tmp/codeartifact.token
RUN npm config set //company-123456.d.codeartifact.us-east-2.amazonaws.com/npm/internal/:_authToken=`cat /tmp/codeartifact.token`
RUN rm -f /tmp/codeartifact.token

And here ARG cannot work and i don't want to use ENV because i don't want to expose this token to anything else