perf (tsort) : avoid reading the whole input into memory and intern strings (#9872)

* perf(tsort): avoid reading the whole input into memory and intern strings

* perf(tsort): avoid redundant check on input

* perf(tsort): switch to the Bucket interning Backend for better lookup performance

Author: sylvestre

Cargo.lock

@@ -1,7 +1,7 @@
 # coreutils (uutils)
 # * see the repository LICENSE, README, and CONTRIBUTING files for more information
 
-# spell-checker:ignore (libs) bigdecimal datetime serde bincode gethostid kqueue libselinux mangen memmap uuhelp startswith constness expl unnested logind cfgs
+# spell-checker:ignore (libs) bigdecimal datetime serde bincode gethostid kqueue libselinux mangen memmap uuhelp startswith constness expl unnested logind cfgs interner
 
 [package]
 name = "coreutils"
@@ -369,6 +369,7 @@ same-file = "1.0.6"
 self_cell = "1.0.4"
 # FIXME we use the exact version because the new 0.5.3 requires an MSRV of 1.88
 selinux = "=0.5.2"
+string-interner = "0.19.0"
 signal-hook = "0.4.1"
 tempfile = "3.15.0"
 terminal_size = "0.4.0"

Cargo.toml

@@ -1,3 +1,4 @@
+#spell-checker:ignore (libs) interner
 [package]
 name = "uu_tsort"
 description = "tsort ~ (uutils) topologically sort input (partially ordered) pairs"
@@ -19,9 +20,11 @@ path = "src/tsort.rs"
 
 [dependencies]
 clap = { workspace = true }
+fluent = { workspace = true }
+string-interner = { workspace = true }
 thiserror = { workspace = true }
+nix = { workspace = true, features = ["fs"] }
 uucore = { workspace = true }
-fluent = { workspace = true }
 
 [[bin]]
 name = "tsort"