This was an oversight. The difference is significant. I had my suspicions about this when I noticed the huge difference in time between writing to /dev/null vs. an actual file during profiling. On one of our systems, here's the number of syscalls _before_ this change: $ strace -c target/release/tameld --emit xmle -o foo foo.xmlo % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 85.05 4.966192 16 318473 write 7.23 0.421977 13 32298 lstat 6.53 0.381424 15 25113 read 0.75 0.043691 13 3350 readlink 0.25 0.014713 61 241 close 0.12 0.007167 30 241 openat 0.05 0.003175 151 21 munmap 0.01 0.000488 14 35 brk 0.01 0.000292 9 33 mmap 0.00 0.000266 38 7 mremap 0.00 0.000004 1 3 sigaltstack 0.00 0.000000 0 6 fstat 0.00 0.000000 0 1 poll 0.00 0.000000 0 11 mprotect 0.00 0.000000 0 7 rt_sigaction 0.00 0.000000 0 1 rt_sigprocmask 0.00 0.000000 0 6 6 access 0.00 0.000000 0 1 execve 0.00 0.000000 0 1 arch_prctl 0.00 0.000000 0 1 sched_getaffinity 0.00 0.000000 0 1 set_tid_address 0.00 0.000000 0 1 set_robust_list 0.00 0.000000 0 2 prlimit64 ------ ----------- ----------- --------- --------- ---------------- 100.00 5.839389 379854 6 total And _after_: $ strace -c target/release/tameld --emit xmle -o foo foo.xmlo % time seconds usecs/call calls errors syscall ------ ----------- ----------- --------- --------- ---------------- 45.21 0.435010 13 32298 lstat 40.09 0.385752 15 25113 read 6.14 0.059113 21 2809 write 4.75 0.045687 14 3350 readlink 2.51 0.024115 100 241 close 0.84 0.008045 33 241 openat 0.26 0.002468 118 21 munmap 0.06 0.000580 17 35 brk 0.06 0.000566 17 33 mmap 0.03 0.000279 40 7 mremap 0.02 0.000181 16 11 mprotect 0.01 0.000087 15 6 6 access 0.01 0.000082 12 7 rt_sigaction 0.01 0.000075 13 6 fstat 0.00 0.000027 9 3 sigaltstack 0.00 0.000024 12 2 prlimit64 0.00 0.000018 18 1 execve 0.00 0.000016 16 1 poll 0.00 0.000013 13 1 sched_getaffinity 0.00 0.000012 12 1 rt_sigprocmask 0.00 0.000012 12 1 arch_prctl 0.00 0.000012 12 1 set_robust_list 0.00 0.000011 11 1 set_tid_address ------ ----------- ----------- --------- --------- ---------------- 100.00 0.962185 64190 6 total What a difference! There's still a lot of other red flags in there; those can be addressed separately. This was originally written as I was learning Rust, and I suspect that I didn't realize that File wasn't buffered at the time. For the above link: times go from 1.23s pre-change to 0.85s after: 0.77user 0.44system 0:01.23elapsed 99%CPU (0avgtext+0avgdata 48520maxresident)k 0inputs+43952outputs (0major+12825minor)pagefaults 0swaps 0.69user 0.15system 0:00.85elapsed 98%CPU (0avgtext+0avgdata 48396maxresident)k 0inputs+43952outputs (0major+12823minor)pagefaults 0swaps |
||
---|---|---|
bin | ||
build-aux | ||
core | ||
design/tpl | ||
doc | ||
progtest | ||
rater | ||
src | ||
tamer | ||
test | ||
tools | ||
.gitignore | ||
.gitlab-ci.yml | ||
.gitmodules | ||
.rev-xmle | ||
.rev-xmlo | ||
COPYING | ||
COPYING.FDL | ||
HACKING | ||
Makefile.am | ||
README.md | ||
RELEASES.md | ||
VERSION.in | ||
bootstrap | ||
c1map.xsd | ||
configure.ac | ||
package-lock.json |
README.md
TAME
TAME is The Algebraic Metalanguage, a programming language and system of tools designed to aid in the development, understanding, and maintenance of systems performing numerous calculations on a complex graph of dependencies, conditions, and a large number of inputs.
This system was developed at Ryan Specialty Group (formerly LoVullo Associates) to handle the complexity of comparative insurance rating systems. It is a domain-specific language (DSL) that itself encourages, through the use of templates, the creation of sub-DSLs. TAME itself is at heart a calculator—processing only numerical input and output—driven by quantifiers as predicates. Calculations and quantifiers are written declaratively without concern for order of execution.
The system has powerful dependency resolution and data flow capabilities.
TAME consists of a macro processor (implementing a metalanguage), numerous compilers for various targets (JavaScript, HTML documentation and debugging environment, LaTeX, and others), linkers, and supporting tools. The input grammar is XML, and the majority of the project (including the macro processor, compilers, and linkers) is written in a combination of XSLT and Rust.
TAMER
Due to performance requirements, this project is currently being reimplemented in Rust. That project can be found in the tamer/ directory.
Documentation
Compiled documentation for the latest release is available via our GitLab mirror, which uses the same build pipeline as we do on our internal GitLab instance. Available formats are:
Getting Started
To get started, make sure Saxon version 9 or later is available and its path
set as SAXON_CP
; that the path to hoxsl is set via HOXSL
; and then run
the bootstrap
script:
$ export SAXON_CP=/path/to/saxon9he.jar
$ export HOXSL=/path/to/hoxsl/root
$ ./boostrap
Running Test Cases
To run the test cases, invoke make check
(or its alias, make test
).
Testing Core Features
In order to run tests located at core/test/core/**
, a supporting environment
is required. (e.g. mega rater). Inside a supporting rater, either check out a
submodule containing the core tests, or temporarily add them into the
submodule.
Build the core test suite summary page using:
$ make rater/core/test/core/suite.html
Visit the summary page in a web browser and click the Calculate Premium button. If all test cases pass, it will yield a value of $1.
Hacking
Information for TAME developers can be found in the file HACKING
.
License
This program is free software: you can redistribute it and/or modify it under the terms of the GNU General Public License as published by the Free Software Foundation, either version 3 of the License, or (at your option) any later version.
This program is distributed in the hope that it will be useful, but WITHOUT ANY WARRANTY; without even the implied warranty of MERCHANTABILITY or FITNESS FOR A PARTICULAR PURPOSE. See the GNU General Public License for more details.