Skip to content

fix: sort_route uses vars count as tiebreaker to match max vars #158

New issue

Have a question about this project? Sign up for a free GitHub account to open an issue and contact its maintainers and the community.

Sign up for GitHub

By clicking “Sign up for GitHub”, you agree to our terms of service and privacy statement. We’ll occasionally send you account related emails.

Already on GitHub? Sign in to your account

Open
janiussyafiq wants to merge 2 commits into
base: master
Choose a base branch
from
Open

fix: sort_route uses vars count as tiebreaker to match max vars #158

Show file tree
Hide file tree
Changes from 1 commit
Commits
Show all changes
2 commits
Select commit Hold shift + click to select a range
0ee4d2f
fix: sort_route uses vars count as tiebreaker to match max vars
janiussyafiq Apr 3, 2026
67e3d5d
fix: used string equality
janiussyafiq Apr 3, 2026
File filter

Filter by extension

Filter by extension
Conversations
Failed to load comments.
Loading
Jump to
Jump to file
Failed to load files.
Loading
Diff view
Diff view
4 changes: 4 additions & 0 deletions lib/resty/radixtree.lua
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -205,6 +205,9 @@ local mt = { __index = _M, __gc = gc_free }

local function sort_route(route_a, route_b)
if route_a.priority == route_b.priority then
if #route_a.path_org == #route_b.path_org then
return (route_a.vars_len or 0) > (route_b.vars_len or 0)
end
return #route_a.path_org > #route_b.path_org
end
Comment on lines 206 to 212
Copy link

Copilot AI Apr 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

The new tiebreaker is triggered when #route_a.path_org == #route_b.path_org, which is broader than the PR’s stated intent (“same priority and uri”). Different paths with the same length (possible in the same routes bucket for param/prefix matches) would now be reordered by vars_len, changing routing precedence beyond identical URIs. If the intent is only to break ties for identical paths, consider checking route_a.path_org == route_b.path_org (string equality) before comparing vars_len, and otherwise keep the existing ordering rules.

Copilot uses AI. Check for mistakes.
Copy link
Copy Markdown
Author

@janiussyafiq janiussyafiq Apr 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

fixed

return (route_a.priority or 0) > (route_b.priority or 0)
Expand Down Expand Up @@ -438,6 +441,7 @@ local function common_route_data(path, route, route_opts, global_opts)
error("failed to handle expression: " .. err, 2)
end
route_opts.vars = route_expr
route_opts.vars_len = #route.vars
end
Comment on lines 434 to 445
Copy link

Copilot AI Apr 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

vars_len is set as #route.vars, but the vars DSL (lua-resty-expr) can include non-condition tokens (e.g., boolean operators like "and"/"or"), so #route.vars may not reflect the number of actual conditions. This can lead to incorrect “more-specific” ordering. Consider computing vars_len as the count of condition nodes (e.g., elements whose type is table) rather than the raw array length, and add a test case that includes boolean operators to lock in the intended behavior.

Copilot uses AI. Check for mistakes.
Copy link
Copy Markdown
Author

@janiussyafiq janiussyafiq Apr 3, 2026

Choose a reason for hiding this comment

The reason will be displayed to describe this comment to others. Learn more.

Valid observation, but over-engineering. The concern is real: a route with vars = {{"AND", cond1, cond2}} gets vars_len = 1 while vars = {cond1, cond2} gets vars_len = 2, even though they express the same thing.
However:

  • In practice, the flat form is what all real users (including APISIX) use. The nested {"AND", ...} grouping syntax is rarely used.
  • Counting actual condition nodes would require recursive traversal and adds meaningful complexity for an edge case.
  • Even with the imprecision, the behavior is still correct — it doesn't break routing, it just has a slight inaccuracy in sort order for an unusual DSL form.
    ps: might need some input from maintainers


local hosts = route.hosts
Expand Down
35 changes: 35 additions & 0 deletions t/vars.t
View file
Open in desktop
Original file line number Diff line number Diff line change
Expand Up @@ -566,3 +566,38 @@ metadata /aa
metadata /aa
nil
nil



=== TEST 20: match max vars condition
--- config
location /t {
content_by_lua_block {
local radix = require("resty.radixtree")
local rx = radix.new({
{
paths = {"/aa"},
metadata = "metadata /aa",
vars = {
{"arg_k", "==", "v"},
},
},
{
paths = {"/aa"},
metadata = "metadata /aa2",
vars = {
{"arg_k", "==", "v"},
{"arg_k2", "==", "v2"},
},
},
})

ngx.say(rx:match("/aa", {vars = ngx.var}))
}
}
--- request
GET /t?k=v&k2=v2
--- no_error_log
[error]
--- response_body
metadata /aa2
Loading