Resolve from the CLI#

Resolve package requirements and export declarations in the terminal

Static preview · VHS tape

Use the one-shot command when conda-presto is installed locally. Save the shared environment.yml and extra-deps.yml fixtures in the current directory:

conda presto -c conda-forge -p linux-64 zlib
conda presto -f environment.yml -p linux-64 -p osx-arm64 \
  > environment.json

The default output is native JSON with a result or error for each platform. Exporter output requires successful solves for every selected platform. The command selects packages without creating or changing a prefix.

To merge both files and an inline requirement into one solve:

conda presto -f environment.yml -f extra-deps.yml -p linux-64 bzip2 \
  > merged.json

The result includes zlib and zstd from the first file, xz from the second, and the inline bzip2 requirement.

To save exact package URLs for a later conda create --file invocation:

conda presto -f environment.yml -p linux-64 --format explicit > explicit.txt

The output starts with @EXPLICIT after any comment headers. Presto writes the package selection without installing it.

To render declared dependencies without solving:

conda presto --export -f environment.yml --format requirements > requirements.txt

This preserves supported declarations without producing a pinned lock. Workspace manifests support selected environment and target export through the same mode, as shown in Parse and solve a workspace manifest.

To convert a covered lockfile without solving or downloading packages:

conda presto -f environment.yml -p linux-64 --format pixi-lock-v6 > pixi.lock
conda presto --export -f pixi.lock -p linux-64 --format conda-lock-v1 > conda-lock.yml

Export mode rejects inputs and selections the output format cannot represent. It does not fall back to a solve. See Export lockfiles without solving for ordinary lock conversion restrictions and Extract a named environment from a saved lock for named workspace lock extraction.

To update a declared direct dependency in one workspace target, keep the manifest unchanged and provide the complete saved lock:

mkdir -p updated
conda presto --update -f conda.lock --manifest conda.toml \
  -e test -p linux-64 numpy > updated/conda.lock
conda presto --validate -f updated/conda.lock --manifest conda.toml

Use the exact environment and logical target declared by your manifest. This updates numpy only in test/linux-64 and preserves saved package selections for every other environment and target. The solver may also change transitive dependencies in the selected target. Presto rejects an inconsistent baseline and writes the complete result only after its final consistency check succeeds. Redirect to a different file so the shell does not truncate your baseline before Presto reads it.

Compare the saved package references for every unselected target:

python - <<'PY'
import json
from pathlib import Path
from conda.common.serialize.yaml import loads

before, after = [loads(Path(name).read_text()) for name in ("conda.lock", "updated/conda.lock")]
for environment, entry in before["environments"].items():
    for target, references in entry["packages"].items():
        if (environment, target) != ("test", "linux-64"):
            updated = after["environments"][environment]["packages"][target]
            references = sorted(json.dumps(ref, sort_keys=True) for ref in references)
            updated = sorted(json.dumps(ref, sort_keys=True) for ref in updated)
            assert references == updated
print("Unselected package references are unchanged")
PY

See CLI reference for flags and exit codes and Output formats for format limitations.

Show Source