<!DOCTYPE html>
<html data-lt-installed="true">
<head>
<meta http-equiv="Content-Type" content="text/html; charset=UTF-8">
</head>
<body style="padding-bottom: 1px;">
<p>Hi, Sergey,</p>
<p>thanks for the patch! See my comments.</p>
<p>Sergey</p>
<div class="moz-cite-prefix">On 10/24/25 14:00, Sergey Kaplun wrote:<br>
</div>
<blockquote type="cite"
cite="mid:edcba097e9269d0814359a276d99f79b4f13ca09.1761301736.git.skaplun@tarantool.org">
<pre wrap="" class="moz-quote-pre">This patch adds a helper script to aggregate the benchmark results from
JSON files to the format parsable by the InfluxDB line protocol [1].</pre>
</blockquote>
<p>format cannot be parsed by protocol, please rephrase.</p>
<p>Something like "the format compatible with the InfluxDB line
protocol"</p>
<blockquote type="cite"
cite="mid:edcba097e9269d0814359a276d99f79b4f13ca09.1761301736.git.skaplun@tarantool.org">
<pre wrap="" class="moz-quote-pre">
All JSON files from each suite in the <perf/output> directory are
considered as the benchmark results and aggregated into the
<perf/output/summary.txt> file that can be posted to the InfluxDB. The
results are aggregated via the new target LuaJIT-perf-aggregate.</pre>
</blockquote>
may be say that cjson is required?
<blockquote type="cite"
cite="mid:edcba097e9269d0814359a276d99f79b4f13ca09.1761301736.git.skaplun@tarantool.org">
<pre wrap="" class="moz-quote-pre">
[1]: <a class="moz-txt-link-freetext" href="https://docs.influxdata.com/influxdb/v2/reference/syntax/line-protocol/">https://docs.influxdata.com/influxdb/v2/reference/syntax/line-protocol/</a>
---
perf/CMakeLists.txt | 13 ++++
perf/helpers/aggregate.lua | 124 +++++++++++++++++++++++++++++++++++++
2 files changed, 137 insertions(+)
create mode 100644 perf/helpers/aggregate.lua
diff --git a/perf/CMakeLists.txt b/perf/CMakeLists.txt
index cc3c312f..68e561fd 100644
--- a/perf/CMakeLists.txt
+++ b/perf/CMakeLists.txt
@@ -97,3 +97,16 @@ add_custom_target(${PROJECT_NAME}-perf
add_custom_target(${PROJECT_NAME}-perf-console
DEPENDS LuaJIT-benches-console
)
+
+set(PERF_SUMMARY ${PERF_OUTPUT_DIR}/summary.txt)
+add_custom_target(${PROJECT_NAME}-perf-aggregate
+ BYPRODUCTS ${PERF_SUMMARY}
+ COMMENT "Aggregate performance test results into ${PERF_SUMMARY}"
+ COMMAND ${CMAKE_COMMAND} -E env
+ LUA_CPATH="${LUA_CPATH}"
+ ${LUAJIT_BINARY} ${CMAKE_CURRENT_SOURCE_DIR}/helpers/aggregate.lua
+ ${PERF_SUMMARY}
+ ${PERF_OUTPUT_DIR}
+ WORKING_DIRECTORY ${CMAKE_CURRENT_BINARY_DIR}
+ DEPENDS luajit-main
+)
diff --git a/perf/helpers/aggregate.lua b/perf/helpers/aggregate.lua
new file mode 100644
index 00000000..12a8ab89
--- /dev/null
+++ b/perf/helpers/aggregate.lua
@@ -0,0 +1,124 @@
+local json = require('cjson')</pre>
</blockquote>
What if cjson is absent? Do we want to handle error?
<blockquote type="cite"
cite="mid:edcba097e9269d0814359a276d99f79b4f13ca09.1761301736.git.skaplun@tarantool.org">
<pre wrap="" class="moz-quote-pre">
+
+-- File to aggregate the benchmark results from JSON files to the
+-- format parsable by the InfluxDB line protocol [1]:
+-- <measurement>,<tag_set> <field_set> <timestamp>
+--
+-- <tag_set> and <field_set> have the following format:
+-- <key1>=<value1>,<key2>=<value2>
+--
+-- The reported tag set is a set of values that can be used for
+-- filtering data (i.e., branch or benchmark name).
+--
+-- luacheck: push no max comment line length
+--
+-- [1]: <a class="moz-txt-link-freetext" href="https://docs.influxdata.com/influxdb/v2/reference/syntax/line-protocol/">https://docs.influxdata.com/influxdb/v2/reference/syntax/line-protocol/</a>
+--
+-- luacheck: pop</pre>
</blockquote>
<p>I propose to document command-line options</p>
<p>(1st arg is output file, 2nd arg is a dir, "current dir by
default"),</p>
<p>env variables (PERF_COMMIT, PERF_BRANCH) and requirements</p>
<p>(git is an optional requirement, cjson Lua module is mandatory).</p>
<blockquote type="cite"
cite="mid:edcba097e9269d0814359a276d99f79b4f13ca09.1761301736.git.skaplun@tarantool.org">
<pre wrap="" class="moz-quote-pre">
+
+local output = assert(arg[1], 'Output file is required as the first argument')
+local input_dir = arg[2] or '.'
+
+local out_fh = assert(io.open(output, 'w+'))
+
+local function exec(cmd)
+ return io.popen(cmd):read('*all'):gsub('%s+$', '')
+end
+
+local commit = os.getenv('PERF_COMMIT') or exec('git rev-parse --short HEAD')
+assert(commit, 'can not determine the commit')
+
+local branch = os.getenv('PERF_BRANCH') or
+ exec('git rev-parse --abbrev-ref HEAD')
+assert(branch, 'can not determine the branch')
+
+-- Not very robust, but OK for our needs.
+local function listdir(path)
+ local handle = io.popen('ls -1 ' .. path)
+
+ local files = {}
+ for file in <a class="moz-txt-link-freetext" href="handle:lines()">handle:lines()</a> do
+ table.insert(files, file)
+ end
+
+ return files
+end
+
+local tag_set = {branch = branch}
+
+local function table_plain_copy(src)
+ local dst = {}
+ for k, v in pairs(src) do
+ dst[k] = v
+ end
+ return dst
+end
+
+local function read_all(file)
+ local fh = assert(io.open(file, 'rb'))
+ local content = <a class="moz-txt-link-freetext" href="fh:read('*all')">fh:read('*all')</a>
+ <a class="moz-txt-link-freetext" href="fh:close()">fh:close()</a>
+ return content
+end
+
+local REPORTED_FIELDS = {
+ 'cpu_time',
+ 'items_per_second',
+ 'iterations',
+ 'real_time',
+}
+
+local function influx_kv(tab)
+ local kv_string = {}
+ for k, v in pairs(tab) do
+ table.insert(kv_string, ('%s=%s'):format(k, v))
+ end
+ return table.concat(kv_string, ',')
+end
+
+local time = os.time()
+local function influx_line(measurement, tags, fields)
+ return ('%s,%s %s %d\n'):format(measurement, influx_kv(tags),
+ influx_kv(fields), time)
+end
+
+for _, suite_name in pairs(listdir(input_dir)) do
+ -- May list the report file, but will be ignored by the
+ -- condition below.
+ local suite_dir = ('%s/%s'):format(input_dir, suite_name)
+ for _, file in pairs(listdir(suite_dir)) do
+ -- Skip files in which we are not interested.
+ if not <a class="moz-txt-link-freetext" href="file:match('%.json$')">file:match('%.json$')</a> then goto continue end
+
+ local data = read_all(('%s/%s'):format(suite_dir, file))
+ local bench_name = <a class="moz-txt-link-freetext" href="file:match('(">file:match('(</a>[^/]+)%.json')
+ local bench_data = json.decode(data)
+ local benchmarks = bench_data.benchmarks
+ local arch = bench_data.context.arch
+ local gc64 = bench_data.context.gc64
+ local jit = bench_data.context.jit
+
+ for _, bench in ipairs(benchmarks) do
+ local full_tag_set = table_plain_copy(tag_set)
+ full_tag_set.name = bench.name
+ full_tag_set.suite = suite_name
+ full_tag_set.arch = arch
+ full_tag_set.gc64 = gc64
+ full_tag_set.jit = jit
+
+ -- Save the commit as a field, since we don't want to filter
+ -- benchmarks by the commit (one point of data).
+ local field_set = {commit = ('"%s"'):format(commit)}
+
+ for _, field in ipairs(REPORTED_FIELDS) do
+ field_set[field] = bench[field]
+ end
+
+ local line = influx_line(bench_name, full_tag_set, field_set)
+ out_<a class="moz-txt-link-freetext" href="fh:write(line)">fh:write(line)</a>
+ end
+ ::continue::
+ end
+end
+
+out_<a class="moz-txt-link-freetext" href="fh:close()">fh:close()</a>
</pre>
</blockquote>
</body>
<lt-container></lt-container>
</html>