replaced HashMap / HashSet with IndexMap / IndexSet (#252)

We discussed making this change before but hadn't got around to it.

`IndexMap` / `IndexSet` has a performance benefit and maintains
consistent ordering (makes testing easier).

Cargo.toml

@@ -88,7 +88,7 @@ serde_json = "1.0.113"
 tokio = { version = "1.36.0", features = ["full"] }
 tokio-util = "0.7.10"
 serde_with = { version = "3.6.0", features = ["base64"] }
-indexmap = { version = "2.2.2", features = ["serde"] }
+indexmap = { version = "2.2.4", features = ["serde"] }
 tempfile = "3.10.0"
 reqwest = { version = "0.11.24", features = ["json", "stream"] }
 futures-util = "0.3.30"

ci/publish.rs

@@ -4,7 +4,7 @@
 //! * `./publish verify` - verify crates can be published to crates.io
 //! * `./publish publish` - actually publish crates to crates.io
 
-use std::collections::HashMap;
+use indexmap::{IndexMap};
 use std::env;
 use std::fs;
 use std::path::{Path, PathBuf};
@@ -64,7 +64,7 @@ fn main() {
         .iter()
         .enumerate()
         .map(|(i, c)| (*c, i))
-        .collect::<HashMap<_, _>>();
+        .collect::<IndexMap<_, _>>();
     crates.sort_by_key(|krate| pos.get(&krate.name[..]));
 
     match &env::args().nth(1).expect("must have one argument")[..] {

crates/api/Cargo.toml

@@ -16,3 +16,4 @@ serde = { workspace = true }
 serde_with = { workspace = true }
 thiserror = { workspace = true }
 itertools = { workspace = true }
+indexmap = { workspace = true }

crates/api/src/v1/content.rs

@@ -2,8 +2,9 @@
 
 pub use super::ContentSource;
 use crate::Status;
+use indexmap::IndexMap;
 use serde::{de::Unexpected, Deserialize, Serialize, Serializer};
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 use thiserror::Error;
 use warg_crypto::hash::AnyHash;
 
@@ -13,7 +14,7 @@ use warg_crypto::hash::AnyHash;
 pub struct ContentSourcesResponse {
     /// The content sources for the requested content digest, as well as additional
     /// content digests imported by the requested content digest.
-    pub content_sources: HashMap<AnyHash, Vec<ContentSource>>,
+    pub content_sources: IndexMap<AnyHash, Vec<ContentSource>>,
 }
 
 /// Represents a content API error.

crates/api/src/v1/fetch.rs

@@ -1,8 +1,9 @@
 //! Types relating to the fetch API.
 
 use crate::Status;
+use indexmap::IndexMap;
 use serde::{de::Unexpected, Deserialize, Serialize, Serializer};
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 use thiserror::Error;
 use warg_crypto::hash::AnyHash;
 use warg_protocol::{
@@ -34,8 +35,8 @@ pub struct FetchLogsRequest<'a> {
     #[serde(skip_serializing_if = "Option::is_none")]
     pub operator: Option<Cow<'a, str>>,
     /// The map of package identifiers to last known fetch token.
-    #[serde(default, skip_serializing_if = "HashMap::is_empty")]
-    pub packages: Cow<'a, HashMap<LogId, Option<String>>>,
+    #[serde(default, skip_serializing_if = "IndexMap::is_empty")]
+    pub packages: Cow<'a, IndexMap<LogId, Option<String>>>,
 }
 
 /// Represents a fetch logs response.
@@ -49,8 +50,8 @@ pub struct FetchLogsResponse {
     #[serde(default, skip_serializing_if = "Vec::is_empty")]
     pub operator: Vec<PublishedRecord>,
     /// The package records appended since last known package record ids.
-    #[serde(default, skip_serializing_if = "HashMap::is_empty")]
-    pub packages: HashMap<LogId, Vec<PublishedRecord>>,
+    #[serde(default, skip_serializing_if = "IndexMap::is_empty")]
+    pub packages: IndexMap<LogId, Vec<PublishedRecord>>,
 }
 
 /// Represents a fetch package names request.
@@ -68,7 +69,7 @@ pub struct FetchPackageNamesRequest<'a> {
 #[serde(rename_all = "camelCase")]
 pub struct FetchPackageNamesResponse {
     /// The log ID hash mapping to a package name. If `None`, the package name cannot be provided.
-    pub packages: HashMap<LogId, Option<PackageName>>,
+    pub packages: IndexMap<LogId, Option<PackageName>>,
 }
 
 /// Represents a fetch API error.

crates/api/src/v1/package.rs

@@ -2,8 +2,9 @@
 
 pub use super::ContentSource;
 use crate::Status;
+use indexmap::IndexMap;
 use serde::{de::Unexpected, Deserialize, Serialize, Serializer};
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 use thiserror::Error;
 use warg_crypto::hash::AnyHash;
 use warg_protocol::{
@@ -25,8 +26,8 @@ pub enum UploadEndpoint {
         url: String,
         /// Optional header names and values for the upload request.
         /// Only `authorization` and `content-type` headers are valid; any other header should be rejected.
-        #[serde(default, skip_serializing_if = "HashMap::is_empty")]
-        headers: HashMap<String, String>,
+        #[serde(default, skip_serializing_if = "IndexMap::is_empty")]
+        headers: IndexMap<String, String>,
     },
 }
 
@@ -50,8 +51,8 @@ pub struct PublishRecordRequest<'a> {
     /// The complete set of content sources for the record.
     ///
     /// A registry may not support specifying content sources directly.
-    #[serde(default, skip_serializing_if = "HashMap::is_empty")]
-    pub content_sources: HashMap<AnyHash, Vec<ContentSource>>,
+    #[serde(default, skip_serializing_if = "IndexMap::is_empty")]
+    pub content_sources: IndexMap<AnyHash, Vec<ContentSource>>,
 }
 
 /// Represents a package record API entity in a registry.
@@ -90,7 +91,7 @@ pub enum PackageRecordState {
     #[serde(rename_all = "camelCase")]
     Sourcing {
         /// The digests of the missing content.
-        missing_content: HashMap<AnyHash, MissingContent>,
+        missing_content: IndexMap<AnyHash, MissingContent>,
     },
     /// The package record is processing.
     #[serde(rename_all = "camelCase")]

crates/client/src/api.rs

@@ -3,13 +3,14 @@
 use anyhow::{anyhow, Result};
 use bytes::Bytes;
 use futures_util::{Stream, TryStreamExt};
+use indexmap::IndexMap;
 use reqwest::{
     header::{HeaderMap, HeaderName, HeaderValue},
     Body, IntoUrl, Method, RequestBuilder, Response, StatusCode,
 };
 use secrecy::{ExposeSecret, Secret};
 use serde::de::DeserializeOwned;
-use std::{borrow::Cow, collections::HashMap};
+use std::borrow::Cow;
 use thiserror::Error;
 use warg_api::v1::{
     content::{ContentError, ContentSourcesResponse},
@@ -237,8 +238,8 @@ impl Client {
     pub async fn latest_checkpoints(
         &self,
         registries: impl Iterator<Item = &String>,
-    ) -> Result<HashMap<String, SerdeEnvelope<TimestampedCheckpoint>>> {
-        let mut timestamps = HashMap::new();
+    ) -> Result<IndexMap<String, SerdeEnvelope<TimestampedCheckpoint>>> {
+        let mut timestamps = IndexMap::new();
         for reg in registries.into_iter() {
             let url = self.url.join(paths::fetch_checkpoint());
             let registry_header = HeaderName::try_from(REGISTRY_HEADER_NAME).unwrap();
@@ -528,7 +529,7 @@ impl Client {
         &self,
         method: &str,
         url: &str,
-        headers: &HashMap<String, String>,
+        headers: &IndexMap<String, String>,
         content: impl Into<Body>,
     ) -> Result<(), ClientError> {
         // Upload URLs may be relative to the registry URL.

crates/client/src/config.rs

@@ -2,11 +2,11 @@
 
 use crate::{ClientError, RegistryUrl};
 use anyhow::{anyhow, Context, Result};
+use indexmap::IndexSet;
 use normpath::PathExt;
 use once_cell::sync::Lazy;
 use serde::{Deserialize, Serialize};
 use std::{
-    collections::HashSet,
     env::current_dir,
     fs::{self, File},
     path::{Component, Path, PathBuf},
@@ -109,7 +109,7 @@ pub struct Config {
 
     /// List of creds availabe in keyring
     #[serde(default, skip_serializing_if = "Option::is_none")]
-    pub keys: Option<HashSet<String>>,
+    pub keys: Option<IndexSet<String>>,
 }
 
 impl Config {