Publishing to crates.io

This workspace contains the main library, the aggregate data crates, and optional per-language dictionary crates named espeak-ng-data-dict-<lang>. Because the main library has optional dependencies on the aggregate data crates, those aggregate crates must be published before espeak-ng.

Publish order

# 1. Core phoneme tables, language definitions, voice definitions (~374 KB)
cargo publish -p espeak-ng-data-phonemes

# 2. Russian language dictionary (~4.5 MB)
cargo publish -p espeak-ng-data-dict-ru

# 3. Optional: publish any language-specific dictionary crates you want to expose
#    individually, for example:
# cargo publish -p espeak-ng-data-dict-en
# cargo publish -p espeak-ng-data-dict-uk

# 4. Aggregate crate for all non-Russian language dictionaries (~3.8 MB)
cargo publish -p espeak-ng-data-dicts

# 5. Main library (after data crates are live on crates.io, remove the
#    [patch.crates-io] section from the root Cargo.toml)
cargo publish -p espeak-ng

Automated publish script

Use the helper script to publish all workspace crates in order:

In --execute mode, the script runs required preflight checks first and aborts publishing if any check fails:

• cargo test
• cargo test --features "c-oracle,bundled-espeak" -- --test-threads=1
• cargo test --features "c-oracle,bundled-espeak" --test oracle_comparison -- --nocapture --test-threads=1

--test-threads=1 is required for the c-oracle checks because libespeak-ng uses global process state; parallel test threads can corrupt it even though the Rust Mutex serialises access at the Rust level.

# Preview the publish order and commands only
python3 scripts/publish_all_crates.py

# Execute publishing for all crates (fails on espeak-ng if [patch.crates-io] is still present)
python3 scripts/publish_all_crates.py --execute

# Execute only data crates (skip main crate)
python3 scripts/publish_all_crates.py --execute --no-main

# Execute with cargo --dry-run for every crate
# Uses `cargo package` (fully offline) instead of `cargo publish --dry-run`
# to avoid crates.io network calls that hang without credentials.
# Skips preflight checks since nothing is actually published.
# Note: [patch.crates-io] is allowed during dry-run.
python3 scripts/publish_all_crates.py --execute --dry-run

# Local dry-run when working tree is not committed
python3 scripts/publish_all_crates.py --execute --dry-run --allow-dirty

# Emergency override: skip preflight checks
python3 scripts/publish_all_crates.py --execute --skip-preflight

Before publishing the main crate

1. Remove the [patch.crates-io] section from the root Cargo.toml.
2. Confirm the version numbers in [dependencies] for the data crates match the versions just published.
3. Run cargo package -p espeak-ng to verify the package is valid.

Aggregate crate sizes (compressed .crate files)

Crate	Files	Compressed
espeak-ng-data-phonemes	406	374 KB
espeak-ng-data-dict-ru	6	4.5 MB
espeak-ng-data-dicts	118	3.8 MB

All aggregate data crates are well under the crates.io 10 MB limit.

Downstream usage

After publication, users add to their Cargo.toml:

[dependencies]
# Core library (text → IPA + synthesis)
espeak-ng = "0.1"

# Optional: embed all data in the binary — no system eSpeak NG needed
espeak-ng = { version = "0.1", features = ["bundled-data"] }

Or, for the split approach (pick only the languages you need):

[dependencies]
espeak-ng               = "0.1"
espeak-ng-data-phonemes = "0.1"   # required for any language
espeak-ng-data-dict-en  = "0.1"
espeak-ng-data-dict-uk  = "0.1"
# espeak-ng-data-dict-ru = "0.1"  # add Russian if needed

Or enable the generated selective features on the main crate:

[dependencies]
espeak-ng = { version = "0.1", features = ["bundled-data-en", "bundled-data-uk"] }

Then in code:

use std::path::PathBuf;

let data_dir = PathBuf::from("/tmp/my-espeak");
std::fs::create_dir_all(&data_dir)?;

// Install the embedded data files once at startup:
espeak_ng_data_phonemes::install(&data_dir)?;
espeak_ng_data_dict_en::install(&data_dir)?;
espeak_ng_data_dict_uk::install(&data_dir)?;

// Or use selective helpers from the main crate:
espeak_ng::install_bundled_languages(&data_dir, &["en", "uk"])?;

// Or use the convenience wrapper (requires bundled-data feature):
espeak_ng::install_bundled_data(&data_dir)?;

let engine = espeak_ng::EspeakNg::with_data_dir("en", &data_dir)?;
let ipa = engine.text_to_phonemes("hello world")?;

Per-language crates are generated from espeak-ng-data/*_dict by scripts/generate_dict_crates.py.