#Prerelease control

How molt cuts prereleases -- and why there is no molt pre mode to enter or exit.

#Molt has no prerelease mode

changesets models prereleases as a persistent state: you run changeset pre enter next, which writes a pre.json file that becomes shared branch state, cut some prereleases, and later run changeset pre exit. That pre.json is the single most-disliked part of changesets, and its arbitrary tags (1.0.1-next.0) are not even legal under PEP 440.

Molt does neither. Prerelease is a stateless flag on molt version:

molt version --pre rc      # cut release candidates this run
molt version               # back to a normal release -- nothing to clean up

There is no pre.json, no mode to enter, no mode to exit, and no leftover state on the branch. The "state" is simply the version string already on disk. This page is the reference for how --pre behaves.

#Synopsis

molt version --pre {a,b,rc,dev} [OPTIONS]

#The --pre values

--pre takes one PEP 440 prerelease identifier. Each maps to a fixed spelling -- arbitrary tags are not allowed:

See Versioning and PEP 440 for the version grammar.

#How the counter moves

Because the counter lives in the version string, not in a state file, iteration is just running the command again:

• First --pre run reads the pending changesets, computes the target bump, and appends the prerelease identifier at counter 0. A pending minor on 1.0.0 with --pre rc yields 1.1.0rc0. The changeset files are kept: they are the counter's input, so a --pre run never consumes them.
• Each subsequent --pre run increments the counter: 1.1.0rc0 to 1.1.0rc1 to 1.1.0rc2. The highest bump type across the accumulated changesets is retained, so a major that landed earlier is not lost when a later minor arrives.
• Exiting prerelease is running plain molt version (no --pre). It finalizes to the stable version, dropping the prerelease identifier: 1.1.0rc1 becomes 1.1.0, and this is the run that consumes the changesets.

Dependents that already opted into prereleases through their own constraints are not force-released as a dependency moves rc0 to rc1 -- this follows PEP 440's opt-in scoping and is a deliberate divergence from changesets. See Dependency propagation and Versioning and PEP 440.

--pre cannot be combined with --snapshot.

#Migrating from changeset pre enter / pre exit

For users coming from changesets, the mapping is direct:

If you have a pre.json from a previous tool, delete it; molt does not read it.

#Examples

Cut a candidate, iterate, then release:

molt version --pre rc      # 1.1.0rc0
# fix something, add a changeset
molt version --pre rc      # 1.1.0rc1
molt version               # 1.1.0

Cut an alpha for early testing:

molt version --pre a       # 1.1.0a0

Preview a prerelease without writing:

molt version --pre rc --dry-run

#See also

• molt version -- the command --pre lives on.
• Prerelease mode -- the concept in depth.
• Migrating from changesets.
• Snapshot releases -- the other throwaway-version path.