Repository navigation
Releases
0.4.0 - 2026-11-19
Release Notes
Added
rift A Bon two directories compares the trees, asdiff -rprints
them: files only in one tree, a unified diff per changed file under a
diff -rheader,Binary files ... differfor files with a NUL in
their first 8 KiB, andRenamed:for a file moved with its content
unchanged. Symbolic links are compared by target and never followed.
--summarylists one status per path and the counts. The whole run
stays under--mem; one directory's listing must fit in a quarter of
it, about 2 million entries at the default256M, or the run is
refused with a limit to rerun with.rift --key --show-fieldsprints under each changed key the fields
that differ, one per line:price: 10 -> 12. Fields are CSV
columns named by the header (numbered forcsv:N), top-level NDJSON
members, orCOPYcolumns. Values are cut at 64 bytes with..., and
(none)marks a field one record lacks.--fields a,bcompares only the named fields and implies
--show-fields: a key whose other fields changed is not printed and
does not count for the exit status.rift-mcphas two tools for key mode.keyed_summarycounts the keys
added, removed and changed between two files and lists the first ten
of each kind.keyed_changesreturns a page of the changed keys with
both records and, for a changed key, the fields that differ, a page at a
time by an opaquecursor: one from another call, or after a file
changed, is refused with a message to start again. A key from
apg_dumpcarries itsCOPYtable as atablefield.keytakes
the CLI'scsv:NAME,csv:N,json:PATHandcopy:NAME, and errors
use the CLI's wording, a budget refusal ending inrerun with --mem.
A keyed run counts itsmemin the server's 512 MiB total and takes
one of its four slots, but is not kept between calls: each page reruns
the whole comparison, and amemunder about 260K to 400K, depending on
the pair, is refused (caching is #89).- Each
.deb,.vsixand.mcpbin a release has a<file>.sha256
next to it, in the format of the archives' checksums:sha256.sum
lists only the filesdistbuilds (#69). bench/compare.sh BASE HEAD SCRATCHbuilds two refs, runs them
interleaved on fixed generated pairs, checks that their outputs match
and prints the medians, spreads, peak heap and load as a table.
Changed
- Output change: with
--key copy:NAME, every key line names its table
before the key,~public.item:1instead of~1, in every output mode,
so the same key in two tables no longer prints as the same line.
Scripts that parseCOPYkey lines must strip the new prefix. A colon
in the table name prints as\:. Duplicate-key errors name the table
too:duplicate keys: public.item:7. CSV and NDJSON output is
unchanged. Inrift-core,KeyedChangegains atablefield and
KeyError::DuplicatelistsTableKeyvalues. - A directory given with a file is refused with
give two files or two directories(exit 2) instead ofnot a regular file. rift servewaits up to 5 seconds for room, at most four calls at a
time, before it answers busy, but only for a client that declares
capabilities.retry_afterininitialize. Other clients keep the
immediate busy answer.initializeignores a capability it does not
know and refuses a known one of the wrong type.
Fixed
--key csv:on a file whose last record leaves a quote open and that
ends in a lone CR reported that record as changed against the same
file without the CR, and kept the line breaks in the key when the
quote was in the key column. The final CR now counts as a line break,
as it does outside quotes, and such files compare equal.- Key mode gave its threads' memory back to the budget before the
threads had exited, so a few KiB they allocated on the way out counted
against no reservation, and underrift serveor the MCP server
against the next comparison. It now waits for them. With--key copy:
the first table name could also take 80 bytes more than reserved. - On a file with duplicate keys,
rift --keycould stop with a
rerun with --memrefusal instead of the duplicate-key error, when a
pass that met the duplicates ended with its share of--memfull.
rift-diff 0.4.0
Install rift-diff 0.4.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.4.0/rift-diff-installer.sh | shDownload rift-diff 0.4.0
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftrift-mcp 0.4.0
Install rift-mcp 0.4.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.4.0/rift-mcp-installer.sh | shDownload rift-mcp 0.4.0
| File | Platform | Checksum |
|---|---|---|
| rift-mcp-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-mcp-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-mcp-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-mcp-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/rift0.3.4 - 2026-11-18
Release Notes
Changed
- The
summaryandnext_differencetools ofrift-mcpsay that their ranges come from the segment map before refinement and can take in equal lines next to a change, and thathunks_atgives the exact lines. rift-mcpwaits up to 5 seconds for room instead of answering busy
at once: a new pair waits for a scan to end or a call to let go of its
diff. At most 4 calls wait at once. A busy answer that still comes
says when to retry: when the nearest scan should end, or in 5000 ms.rift servestill answersbusyat once, and its error data now
carriesretry_after_ms: when the nearest scan should end, or 50 ms.
Waiting likerift-mcpwaits for the VS Code extension to honour it
(#64).
Added
- Each release carries
rift-mcp-<version>.mcpb, an MCP bundle of
rift-mcpfor macOS (universal) and Linux (x86_64, aarch64), with the
same provenance files as the archives, and publishes it to the MCP
Registry asio.github.rubtsov-stan/rift-mcp. MCPB clients install it
from the file. bench/linux-safe.sh --suite toolsruns Rift next to GNUdiff,
git diff --no-index, difftastic, DuckDB, csvdiff, qsv and daff on the
same generated pairs, and keeps the counts each tool reports next to
Rift's.
rift-diff 0.3.4
Install rift-diff 0.3.4
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.4/rift-diff-installer.sh | shDownload rift-diff 0.3.4
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftrift-mcp 0.3.4
Install rift-mcp 0.3.4
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.4/rift-mcp-installer.sh | shDownload rift-mcp 0.3.4
| File | Platform | Checksum |
|---|---|---|
| rift-mcp-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-mcp-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-mcp-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-mcp-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/rift0.3.3 - 2026-11-14
Release Notes
Changed
- Files with a few very long lines (e.g. 16 lines of 64 MiB) can diff
up to ~12% slower than 0.3.2; 0.3.2 was faster there only because its
checkpoints depended on the thread count. Byte-position checkpoints
are planned for 0.4. rift servereportsmoves: "partial"insummaryandmoveswhen the
block list filled up and later moves may be missing; it saidfound.
Clients must toleratemovesvalues they do not know. The VS Code
extension showspartialand reads an unknown value asfoundwhen blocks are listed.- The
--mema refusal advises is at least twice the--memthat was
refused, in the CLI,rift serveand the MCP server, so following it
reaches a working budget in a few reruns instead of many small steps
(a nokey 4 MiB pair from 64K took 20 refusals, now 5). It can
overshoot the real need by up to a factor of two. --statsnames a--memto try (twice the one used) next to the
segments that came out approximate for lack of budget. On the shuffled
16 MiB and nokey 4 MiB pairs, every budget in 32K steps from 10300K to 11800K that
showed such segments is clear at the figure named.- Documented that a full diff needs about 10.3 MiB (
--mem 10475K)
whatever the files hold, in--help, the man page and the memory
guide. - Documented that a moved block is an exact copy that may be partly
aligned: lines the diff aligned with another copy print as context. - Just above the full diff's smallest
--mem, more regions can come out
approximate than in 0.3.2, and the diff can skip moved blocks 0.3.2
listed: the block list is now set aside for one block per seed before
the workers, with or without--no-moves. On the pairs measured this
ended within 1.2 MiB of the smallest--mem(0.45 MiB on a shuffled
pair, 1.13 MiB on a nokey pair); from there up the output matched
0.3.2. - The output is identical on any machine for the same
--memand
input. A run sizes its work for one thread per MiB of--mem, from 1
to 8, in each of the scan, the refinement and the shared pool, and
starts no more threads than the machine has cores;
RAYON_NUM_THREADScan make the shared pool smaller. The cores and
RAYON_NUM_THREADSnow change only the speed. The reserve kept out of the budget is 48 KiB plus 12 KiB
per thread that--memallows, so it no longer grows with the cores.
Below 8 MiB, machines with more cores than MiB of--memrun fewer
threads than before. On more than 8 cores, the diff and--summary
run at most 8 scan threads, 8 refine workers and 8 threads in the
shared pool. - Refusal points moved with the reserve. Measured on an Apple M1 with 8
cores, where the reserve at 8 MiB and up is unchanged: the full diff
of two one-line files still needs 10475K,--summaryof them needs
285K instead of 369K.--summaryof a generated 1 GiB SQL pair needs
1874K instead of 3020K: a scan cut into many chunks no longer runs out
of room for checkpoints. On a machine with fewer cores the reserve at
8 MiB and up grows by 12 KiB per missing core, so the same full diff
needs up to 84K more than in 0.3.2 (10391K to 10475K on one core). - Documented that Rift does not detect binary files: NUL bytes are
printed in the diff and there is noBinary files differmessage. - Key mode's output order depends on the files and
--memonly. The
same pair under the same limit prints the same bytes on every run, on
any machine and any number of threads. The order at a given--mem
can differ from 0.3.1's where key mode used several threads; the set
of lines is the same. - Key mode runs one shard per MiB of the refine half, from 1 to 8, on a
thread pool of its own with one thread per shard, at most one per core
orRAYON_NUM_THREADS. It uses no more than 8 threads on any machine. - Key mode runs at most 64 passes. A
--memunder which the files
would take more, counted from the records read so far, is refused with
a budget error instead of rereading both files hundreds of times. The
error names a--memat least about twice as large; following it
reaches one that works in a few reruns. - Key mode keeps 48 KiB of
--memout of the budget, without the 12 KiB
per pool thread: it no longer starts the shared pool, and holds its
own pool's heap, 16 KiB per shard, in the budget.
Fixed
- A gap whose two sides share no line, reached after 32 nested anchor
levels, is replaced exactly instead of being marked approximate. rift-mcprenders a cached hunk without copying its edit script, so a
call no longer raises the heap by the size of the script.- An error that names a path longer than 256 bytes shows its first and
last 120 bytes, so a command line of 200 KB no longer prints back whole.
The command line is still outside the budget. - A seek hint at the end of a file without a final newline no longer
makes a read of an earlier line return the wrong bytes. No caller
passed such a hint. - Near the full diff's smallest
--mem,--summarylisted moved blocks
that the full diff under the same--memsaid it had not searched.
Where the full diff runs,--summarynow looks for moves exactly when
it does; below the full diff's smallest--memit still lists moves
when the search fits. Within a few tens of KiB above that point the
full diff can now skip moves that 0.3.2 listed, since the block list is
reserved for one block per seed before the search. - Two moved blocks could share lines of B, and
--summarycounted
those lines twice. A block that starts inside another's B range now
loses the shared lines, and is dropped when fewer than 3 remain. - Looking for moved blocks could change the diff: the block list took
its room from the refine half before the scheduler, so a region could
be approximate for one reason with the search and another without it.
The list's room is now sized from--memand the seeds alone and set
aside before the workers, with or without--no-moves. So a region
that 0.3.2 refined exactly under--no-movescan now come out
approximate near the same--mem, as it does with the search on. - A move search whose list filled up dropped every block and reported
moved not searched, the budget is too small. It now keeps the
blocks found so far and says the list is full and later moves are not
listed. The list holds about one block per 512 bytes of--mem. - The full diff,
--summaryand--statscould print different, equally
valid output under different thread counts: each thread of the pool
took 12 KiB from the budget, which moved the stratum or the point where
a region came out approximate (300 000 equal lines with one changed, at
--mem 192000K: 11 lines with one thread, 600 000 lines with 64). - The full diff and
rift serveset aside refine workers' buffers for
as many workers as half of the free refine half holds, at most 8 (4
for a session ofrift serveorrift-mcp, which starts at most 4),
however many start. The scheduler's share used to shrink with each
worker, so a machine with more cores could mark a region approximate
where one with fewer cores printed it exactly. Machines with more than
8 cores start at most 8 refine workers. - The line index's size no longer depends on the scan's chunk size,
which followed the core count. Near some--memvalues a few bytes of
difference picked another stratum, so two machines printed different,
equally valid diffs. The index now keeps one checkpoint per multiple
of its interval, so two checkpoints can be up to twice the interval
apart. - Key mode: a duplicate key could be dropped without an error under a
tight--mem; it is now always an error. With no room left to list
the duplicate, 0.3.2 and earlier skipped the second record and could
report the key as removed or changed. Such a run now ends in a budget
error whose--memhas room to list the key, and a rerun there names
the duplicate. - Key mode names duplicate keys instead of refusing the budget once a
pass has met them. A refusal can still come first when the duplicates
lie in a part of the hash range no pass has reached; following its
advice reaches the duplicate-key error. rift servereports a segment whose refinement ended approximate as
approxinsegmentspages, also when nohunkwas asked for it.
Before, a page showed only the map's status, so a client that re-read
pages after refinement never saw the mark.- A scan cancelled just as its reading ends no longer runs the move
search, which reads both files again, before it stops. - The protocol document says that an
openright after closing a
scanning session can answerbusyuntil the scan has stopped, and
that the client retries. - A
not_readyerror on a failed session says "session 1 failed; open
the files again" instead of asking the client to wait for it. rift serveansweredbusyto anopenthat needed a ready session's
room while the notifier happened to hold that session for a look. The
opennow waits up to 250 ms for the notifier to let go, and never
evicts a session whose memory it cannot free.rift servecould sendprogressorsession_statefor a session
after it had answered that session'sclose. The notifier now lists
the live sessions under the lockcloseanswers under.rift servecould send a scan's lastprogressnotification short of
its total and thensession_stateready, when the scan ended between
the two reads. The notifier reads the state first.rift serveunder a small--max-total-memno longer advertises a
default_memormax_session_memabove what its sessions can hold.
Everyopenused to answerbusyforever; one that asks for more than
the total now getstoo_large. The server refuses to start when its
limits leave a session less than 11 MiB, the least that runs even a
one-line pair...
0.3.2 - 2026-11-10
Release Notes
Fixed
- Earlier versions read FIFOs and process substitution (for example
rift <(cmd) file) as empty files and printed a wrong diff with no
error. Rift now refuses any operand that is not a regular file, with the
path and a message to write the input to a file first. A directory names
its path too, and-is documented as a file name. - A Linux
/procfile, which reports a length of 0 but has content, is
refused the same way instead of being read as empty. - A file cut while Rift reads it, or one that holds less than it reports,
is an error naming the file. Key mode used to take the early end as the
end of its records and print a wrong diff with no error. - A file that grows while Rift reads it no longer yields bytes past the
length taken at open, which could mix two versions of the file in one
diff. Reads stop at that length. rift serveno longer hangs when a path is swapped for a FIFO between
its check and itsopen. The open does not block, refuses a file that
is not regular and never follows a symbolic link.rift git-diffno longer blocks on a FIFO operand while it looks for a
Git LFS pointer.- A source that returns short reads could make the line reader return the
tail of a long line as a whole line instead of refusing it as too long.
The reader now keeps the line's start across short reads.
Known issues
- The two known issues of 0.3.1 remain in this release: key mode's line
order can differ between runs once it uses several threads, and key mode
at exactly--mem 16Mruns on one thread. Both are fixed onmainfor
the next release.
rift-diff 0.3.2
Install rift-diff 0.3.2
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.2/rift-diff-installer.sh | shDownload rift-diff 0.3.2
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftrift-mcp 0.3.2
Install rift-mcp 0.3.2
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.2/rift-mcp-installer.sh | shDownload rift-mcp 0.3.2
| File | Platform | Checksum |
|---|---|---|
| rift-mcp-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-mcp-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-mcp-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-mcp-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/rift0.3.1 - 2026-11-09
Release Notes
Added
rift --key --show-recordsprints the records under each change: A's
behind-, B's behind+, with the prefix on each line of a
record that spans lines. Rift reads the records back from the files
through a 64 KiB buffer taken from the--membudget.--statsprints the peak heap after an error as well, and
rift git-diff --statsprints it after the diff or the error.
Changed
--memandrift git-diff --memare at least 64K. Below that, parsing
the command line alone took more heap than the limit. The reserve
leaves key mode less room at a small--mem, so it takes more passes:
at 1M a 128 MiB CSV pair takes 145 passes instead of 122 and runs about
19% longer.- With the reserve below, a run that only just fit a small limit can now
stop with a budget error.--summaryof a 32 MiB SQL dump at
--mem 1Mdid on 8 threads. Thererun with --mem Nit prints names a
limit that gets past that point; here 1089K, which gives the same
summary as before. - An approximate hunk says why in plain words: the unified diff writes
@@ ... @@ approx: <reason>,--statscounts the segments for each
reason, the terminal view shows the reason in the status line when it
fits, andrift-mcpadds it to the hunk header.Hunk::approxin
rift-coreis now anOption<ApproxReason>.
Fixed
- Key mode at a small
--mempeaked up to 10 KiB over the limit: its
output buffers and other heap outside the budget went uncounted, and in
the other modes so did the thread pool's heap. Rift now keeps 48 KiB
plus 12 KiB per pool thread (one per core, orRAYON_NUM_THREADS) of
--memout of the budget for that heap, and
rerun with --mem Nadds it to its advice. - Output closed early, as in
rift a b | head, no longer prints an error.
Rift exits 141 without a message, the status a shell reports for
diffkilled by SIGPIPE. rift git-diffexits 0 without a message when git's pager quits early,
sogit diffno longer reports "external diff died".- A failed write to standard output says "write failed", not "read failed".
- Two files, or two regions, that share no line are no longer marked
approx. Deleting one side and inserting the other is the smallest
diff there, and Rift now proves it from the line hashes: a different
hash means a different line. The script is the same as before. --summaryno longer prints anapproxline. It refines nothing, so
its count was always 0 whatever the files.- The
summaryofrift serveandrift-mcpcountedapproxsegments
by the map, which never marks one, so it read 0.approxnow counts
the refined segments that came out approximate, anddiffersthe other
differing segments; together they are still every differing segment.
rift-mcpaddsrefined_segments; inrift serve,progress.doneis
that count.
Known issues
- Key mode at exactly
--mem 16Mruns on one thread: the reserve takes
the refine half just under the 8 MiB key mode needs for several. On a
128 MiB CSV pair it takes 2.21 s instead of 1.25 s in 0.3.0, about 77%
longer.--mem 16528Kor more restores the old time. - Key mode's line order can differ between runs once it uses several
threads. The set of lines is always the same.
rift-diff 0.3.1
Install rift-diff 0.3.1
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.1/rift-diff-installer.sh | shDownload rift-diff 0.3.1
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftrift-mcp 0.3.1
Install rift-mcp 0.3.1
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.1/rift-mcp-installer.sh | shDownload rift-mcp 0.3.1
| File | Platform | Checksum |
|---|---|---|
| rift-mcp-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-mcp-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-mcp-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-mcp-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/rift0.3.0 - 2026-11-08
Release Notes
Added
rift git-diff, the command git runs for an external diff. It takes
the seven arguments git passes and prints a unified diff with git's
headers, and reads a Git LFS object in place when a side is a pointer.rift serve, Rift's protocol on standard input and output, JSON-RPC 2.0
framed as in LSP, for editors.docs/protocol.mdspecifies it.rift-session, the crate that keeps the live diffs shared by
rift serveandrift-mcp.- A VS Code extension in
editors/vscode, built onrift serve. It is not
yet published.
Changed
serveandgit-diffare subcommands, listed inrift --help, so a
file namedserveorgit-diffmust now be given as./serveor
./git-diff.
rift-diff 0.3.0
Install rift-diff 0.3.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.0/rift-diff-installer.sh | shDownload rift-diff 0.3.0
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftrift-mcp 0.3.0
Install rift-mcp 0.3.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.3.0/rift-mcp-installer.sh | shDownload rift-mcp 0.3.0
| File | Platform | Checksum |
|---|---|---|
| rift-mcp-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-mcp-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-mcp-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-mcp-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |
Verifying GitHub Artifact Attestations
The artifacts in this release have attestations generated with GitHub Artifact Attestations. These can be verified by using the GitHub CLI:
gh attestation verify <file-path of downloaded artifact> --repo rubtsov-stan/riftYou can also download the attestation from GitHub and verify against that directly:
gh attestation verify <file-path of downloaded artifact> --bundle <file-path of downloaded attestation> --repo rubtsov-stan/riftVSIX files rebuilt on 2026-11-08 to update the store README; binaries unchanged.
VSIX files rebuilt again on 2026-11-08 to rename the extension to rubtsov-stan.rift-diff, since the Marketplace already has another extension named rift-vscode; binaries unchanged.
v0.2.0
Added
- Shingle anchors: where a region has no line that is unique in both
files, Rift now anchors on short runs of lines, so fewer regions come
out as approximate replacements. --key SPECcompares records by key and ignores row order.SPECis
csv:N,csv:NAME,json:PATHorcopy:NAME(the COPY blocks of a
pg_dumpfile). It prints added, removed and changed keys with their
line numbers, compares record bytes before it calls a record unchanged,
and stops with exit status 2 on a duplicate key. A larger--memmeans
fewer passes over the files.- Moved blocks:
--summaryand--statslist blocks of at least three
lines that appear in both files at different places, and the terminal
view marks both ends of a block and jumps from one end to the other with
m. The unified diff does not change. Detection starts from sampled
lines and does not find every block. --no-movesskips the search for moved blocks. The unified diff is the
same;--summary,--statsand the terminal view say that detection
was off. It cannot be combined with--key.rift-mcp, an MCP server that lets a client explore the diff of two
large files withsummary,hunks_atandnext_difference. It is
released as its own archives for the four targets and with its own shell
installer.- The terminal interface is an optional
tuifeature, on by default. In a
build without it,-iandrift tuiexit with status 2. .debpackages for x86_64 and aarch64 Linux, attached to each release.- A man page,
man/rift.1, generated from the command-line definition
withcargo xtask manand included in theriftrelease archives and
the.debpackages. bench/linux-safe.shruns the benchmark on a Linux host from RAM only,
with memory caps, so it leaves the host and its disks alone.- Fuzz targets for the line scanner, the key parsers, the pipeline, the
unified writer, the terminal sanitizer and the MCP tools, with
fuzz/run-linux.shto run them on a Linux host. - Release archives carry a build provenance attestation that
gh attestation verifychecks. - CI runs
rift-coretests under Miri,cargo-denyandcargo-audit,
CodeQL and an OpenSSF Scorecard check. Dependabot watches cargo and
GitHub Actions, and third-party actions are pinned by commit.
Changed
--statsand--summaryprint the peak heap as
peak heap X of Y (N bytes): the rounded figure for reading and the
exact byte count for checking against--mem.- In key mode, an error after some changes were printed now says on
standard error that the output above is incomplete and must be
discarded. Check the exit status before using key output. - A budget error ends with a
--memto rerun with, in every mode:
refine budget exceeded: requested 5290144 bytes, 5242880 available; rerun with --mem 10333K. It is not a lower bound: a smaller limit can
take another path and succeed. A run at that limit can stop at a later
stage and name a higher one. A diff whose refine half cannot hold one
worker reports the whole amount at once. In key mode, a record too
long for the budget is read to its end, and the limit named is one
under which it fits. - There is no fixed minimum
--mem. The 0.1.0 notes said a budget below
about 12 MiB is refused; the smallest budget that works depends on the
files. - Key-mode errors about a missing column or key name the file as well
as the side:a (data/a.csv) has no column "zz". - Budget errors start in lowercase (
skeleton budget exceeded), and an
I/O error shows its text once instead of twice. - Faster scanning, anchor selection, chain search, scheduling, unified
output and key mode on large files, with the same output. DESIGN.md
lists what changed and how it was measured. release.ymlis generated by cargo-dist and changes only by upgrading
cargo-dist and regenerating it; Dependabot no longer updates it.
Fixed
- Header timestamps at the ends of the time range no longer overflow.
-iandrift tuiexit with status 2 and-i needs a terminalwhen
standard input or output is not a terminal, instead of panicking.- Key mode no longer reports the last CSV record as changed when a quote
left open runs to the end of the file and the files differ only in the
final newline or its line ending (7dd49f5). - The same holds when such a file ends in blank lines that differ only in
number or line endings (b6b1de8). - Key mode on a
pg_dumpfile with many tables no longer slows down and
uses more memory with more threads: chunks parsed in parallel could
register tables a sequential read never sees. The output was not
affected (3696ddd).
Known limits
- Speed on very low-entropy input (files of a few distinct short lines)
is not yet characterised; memory stays within--mem.
v0.1.0
Install rift-diff 0.1.0
Install prebuilt binaries via shell script
curl --proto '=https' --tlsv1.2 -LsSf https://github.com/rubtsov-stan/rift/releases/download/v0.1.0/rift-diff-installer.sh | shDownload rift-diff 0.1.0
| File | Platform | Checksum |
|---|---|---|
| rift-diff-aarch64-apple-darwin.tar.xz | Apple Silicon macOS | checksum |
| rift-diff-x86_64-apple-darwin.tar.xz | Intel macOS | checksum |
| rift-diff-aarch64-unknown-linux-gnu.tar.xz | ARM64 Linux | checksum |
| rift-diff-x86_64-unknown-linux-gnu.tar.xz | x64 Linux | checksum |