Hey folks, how should I set up a python target that has a build step? Specifically, I'm using the Prisma Python Client (

prisma

pypi package). Coming from Poetry, I need to run

prisma generate

in my venv after

poetry install

.

prisma generate

reads the prisma schema, generates python classes, and writes the generated code to the

prisma

site-packages folder. Then application code consumes the generated code from the site package as

from prisma import models

.

Hmm, interesting question. I think writing to the site-packages folder of the venv is… unexpected behaviour for a code generation step. Pants generally supports code gen where it becomes a “normal” file, more like if it was manually placed adjacent to the normal code. Does prisma have options to change how the codegen works?

Yes I can generate to somewhere else, in my repo. I am still not sure how to proceed pants-wise in that case

Ah cool. One way would be using https://www.pantsbuild.org/2.19/docs/ad-hoc-tools/integrating-new-tools-without-plugins where the

runnable

is the prisma requirement target (something like

pants list :: | grep prisma

might find it) and you load up

execution_dependencies

with the code required as input, plus

output_files

pointing to the files that need to be exported. You’ll need to wrap this up in https://www.pantsbuild.org/2.19/reference/targets/experimental_wrap_as_python_sources after

cool, I will try the adhoc approach first. thanks!

Hello I am sorry to reopen a very old thread. I have a similar issue and I trie de to apply the same solution. I have a service whose code is in a folder

services/foo-service

, inside services/foo-service, I have a file

db/schema.prisma

and the package

foo_service

. Previously, I was using default

prisma

generation and now I tried to modify so that the generated client would be in

foo_service/prisma

:

# modified services/foo-service/db/schema.prisma to generate in foo_service
generator client {
  provider             = "prisma-client-py"
  recursive_type_depth = 5
  output               = "../foo_service/prisma"
}

# modified services/foo-service/BUILD file
python_sources()

files(name="db-schema", sources=["db/schema.prisma", "db/migrations/*"])


adhoc_tool(
    name="generate-db-schema",
    runnable="//:reqs#prisma",
    execution_dependencies=[":db-schema", "//:reqs#prisma"],
    args=["generate", "--schema=./db/schema.prisma"],
    output_directories=["foo_service/prisma"],
    root_output_directory=".",
    outputs_match_mode="all",
    log_output=True,
    workdir=".",
)

experimental_wrap_as_python_sources(
    name = "prisma_client",
    inputs = [":generate-db-schema"],

    # Include all files in sources. The generator produces python files and the
    # `schema.prisma` file.
    outputs = ["**"],
)

# modified python code using prisma in foo_service package
from foo_service.prisma import ...

# try to use the `prisma_client` in BUILD at the root of my repo
adhoc_tool(
    name="generate-schema",
    runnable="//services/foo-service/foo_service/custom_openapi_generator.py",
    args=["openapi.json"],
    execution_dependencies=[
        "//services/foo-service:prisma_client",
        "//:reqs",
    ],
    output_files=["openapi.json"],
    root_output_directory=".",
)

And I get the message:

ImportError: cannot import name 'Prisma' from 'foo_service.prisma

. In the sandbox, prisma code is copied at the root of the sandbox in

foo_service/prisma

folder... Thanks for any help 🙏

What action/goal/command are you doing that gives that error message?

I am executing pants test on a unit test in my service. I am a new user of pants and trying to use it as a build system for our Python mono repo.

Cool, thanks for the info. • When you say "in the sandbox", that's the sandbox for that unit test run? • Where's the rest of the "real" code in that sandbox? I.e what's the directory structure?

• When you say "in the sandbox", that's the sandbox for that unit test run?

Yes, it is an integration test, and here what the result of

ls

on that sandbox folder:

ls pants-sandbox-*    
pants-sandbox-ExuorW:
__run.sh			custom_openapi_generator.pex_bin_python_shim.sh	foo_service		openapi.json			python-default.lock
custom_openapi_generator.pex				custom_openapi_generator.pex_pex_shim.sh		modules				pants-requirements.txt		services

• Where's the rest of the "real" code in that sandbox? I.e what's the directory structure?

foo_service

subfolder contains the

prisma

generated code.

services

subfolder contains the folder hierarchy with only the related real code

services/foo-services/foo_service/...

.

okay, I think you might want to try to get the new code generated within that same subfolder

Yes, that's already what I tried. But now I removed

root_output_directory="."

and

workdir="."

and it worked better. The code is now copied in the right place

services/foo-service/foo_service/prisma

.

# modified services/foo-service/BUILD file
python_sources()

files(name="db-schema", sources=["db/schema.prisma", "db/migrations/*"])


adhoc_tool(
    name="generate-db-schema",
    runnable="//:reqs#prisma",
    execution_dependencies=[":db-schema", "//:reqs#prisma"],
    args=["generate", "--schema=./db/schema.prisma"],
    output_directories=["foo_service/prisma"],
    # root_output_directory=".",
    outputs_match_mode="all",
    log_output=True,
    # workdir=".",
)

experimental_wrap_as_python_sources(
    name = "prisma_client",
    inputs = [":generate-db-schema"],

    # Include all files in sources. The generator produces python files and the
    # `schema.prisma` file.
    outputs = ["**"],
)

Now, I have another issue with explicit requirements: if you import a 3rd party dependency and this dependency is not explicitly mentionned in the requirements, even if it is in the lock file, it will not be found by Pants. But this is definitely another issue. Thanks for the help

That's a dependency for running the code-generation, but not a dependency of the generated code. Potentially

output_dependencies=["//:reqs#tomlkit"]

is what you need (https://www.pantsbuild.org/prerelease/reference/targets/adhoc_tool#output_dependencies)

I tried that but it is not working either: in

$ROOT_FOLDER/services/foo-service/BUILD

I have:

files(name="db-schema", sources=["db/schema.prisma", "db/migrations/*"])


adhoc_tool(
    name="generate-db-schema",
    runnable="//:reqs#prisma",
    execution_dependencies=[":db-schema", "//:reqs#prisma"],
    output_dependencies=["//:reqs#tomlkit"],
    args=["generate", "--schema=./db/schema.prisma"],
    output_directories=["foo_service/prisma"],
    outputs_match_mode="all",
    log_output=True,
)

experimental_wrap_as_python_sources(
    name = "prisma_client",
    inputs = [":generate-db-schema"],

    # Include all files in sources. The generator produces python files and the
    # `schema.prisma` file.
    outputs = ["**"],
)

Then in

$ROOT_FOLDER/BUILD

, I have

adhoc_tool(
    name="generate-schema",
    runnable="//services/inference/inference_service/controller/api/export_openapi.py",
    args=["openapi.json"],
    execution_dependencies=[
        "//services/inference:prisma_client",
        "//:reqs",
    ],
    output_files=["openapi.json"],
    root_output_directory=".",
)

run_shell_command(
    name="run-generate-schema",
    execution_dependencies=[":generate-schema"],
    command="exit 0",  # if you just want to run shell command
)

When I run

pants run :run-generate-schema

it failed with the same error. When I run

pants dependencies :generate-schema

I get:

//:reqs#tomlkit
services/inference:prisma_client

I'm not sure sorry. If you can create a standalone reduced example that demonstrates your problem and you can share, it'll be easier for me to help you.