Sometimes you might want to give browscap a file that doesn't live on the filesystem (say, for tests, or because you want to give it a compressed file). This commit lets you do that by passing in any buffered-file object, rather than requiring a filename.

Hello, here is the sample code that gives wrong result on a latest ini file

package main
import (
    "fmt"
    bgo "github.com/digitalcrab/browscap_go"
)
func main() {
    if err := bgo.InitBrowsCap("browscap_min.ini", false); err != nil {
        panic(err)
    }
    browser, ok := bgo.GetBrowser("Mozilla/5.0 (iPad; CPU OS 5_0_1 like Mac OS X) AppleWebKit/534.46 (KHTML, like Gecko) Version/5.1 Mobile/9A405 Safari/7534.48.3")
    if !ok || browser == nil {
        panic("Browser not found")
    } else {
        fmt.Printf("Browser = %s [%s] v%s\n", browser.Browser, browser.BrowserType, browser.BrowserVersion)
        fmt.Printf("Platform = %s v%s\n", browser.Platform, browser.PlatformVersion)
        fmt.Printf("Device_Type = [%s]\n", browser.DeviceType)
        fmt.Printf("IsCrawler = %t\n", browser.IsCrawler())
        fmt.Printf("IsMobile = %t\n", browser.IsMobile())
    }
}

Device_Type = [Mobile Device] This is wrong The right is from - http://browscap.org/ua-lookup gives device_type - Tablet

We're using browscap in a real time application so the original application was too slow. In this pull request I rewrote a part of re0, added tree search for expressions, added browser object caching and added a couple of new attributes to the browser object.

I also added two benchmark tests to compare performance. Initialization is around 30% slower because it has to build the search tree, but the GetBrowser function is 140 times faster.

Speed and memory usage comparison:

HEAD: BenchmarkInit-4 2 758992121 ns/op 165539440 B/op 2681838 allocs/op BenchmarkGetBrowser-4 200 8535325 ns/op 3250 B/op 15 allocs/op

New version: BenchmarkInit-4 1 1163491977 ns/op 136829088 B/op 3190900 allocs/op BenchmarkGetBrowser-4 20000 62218 ns/op 32 B/op 1 allocs/op

they are actually pretty different in some cases and both have useful information.

typical example: Device_Name="Galaxy S5 LTE" Device_Code_Name="SM-G900F"

Fixes https://github.com/digitalcrab/browscap_go/issues/4

I also updated test-data with the newest version of rules so I was able to write a test for the reported issue.

Sorry I didn't see the issue earlier.

https://github.com/browscap/browscap/wiki/Specification:-Lookup-Algorithm says:

If there are multiple matching patterns the longest pattern should be used to identify the browser. If there are multiple patterns of the longest length then the pattern that is earliest in the list of patterns should be used to identify the browser.

The old code, as far as I can tell -- which isn't very far, to be fair -- always picked the earliest in the list of patterns, regardless of length. I did the minimal changes to implement the algorithm as specified. This possibly saved some space in the bargain, as I lowered the space required for scores from 64 bits to 32.

I added a test-case that failed with the old code but passes with the new. It requires an up-to-date browscap.ini file, so I updated the one in test-data as part of this diff.

Before, it was only looking at Device_Code_Name. But Device_Name is a separate field that sometimes has a different value than Device_Code_Name. This change makes it so we prefer the former to the latter.

This is an API-breaking change, but I think it's the right thing to do based on my (admittedly scant) understanding of these ini fields.

UA: Mozilla/5.0 (Windows NT 6.3; Win64; x64; rv:56.0) Gecko/20100101 Firefox/56.0

ini version

[GJK_Browscap_Version]
Version=6000031
Released=Thu, 29 Nov 2018 08:50:14 +0000
Format=php
Type=FULL

result

Browser = Firefox [Browser] v0.0
Platform = Win8.1 v6.3
Device = Windows Desktop [Desktop] unknown
IsCrawler = false
IsMobile = false

browscap result

browser_name_regex | /^mozilla\/5\.0 \(.*windows nt 6\.3.*win64. x64.*\) gecko.* firefox\/56\.0.*$/
browser_name_pattern | mozilla/5.0 (*windows nt 6.3*win64? x64*) gecko* firefox/56.0*
parent | Firefox 56.0
comment | Firefox 56.0
browser | Firefox
browser_type | Browser
browser_bits | 64
browser_maker | Mozilla Foundation
browser_modus | unknown
version | 56.0
majorver | 56

This pull request fixes https://github.com/digitalcrab/browscap_go/issues/6 and fixes https://github.com/digitalcrab/browscap_go/issues/8

Unfortunately I didn't notice the issues earlier because it wasn't an issue with older versions of browscap data. I used to score lines by length but switched to line number because some crawlers weren't detected correctly in old versions of data if I did it with the rule length so I just assumed that the order of lines is the correct logic.

Plus some minor optimizations and fixes.

• Add an option to initialize the browscap library with a Reader, instead of strictly from an .ini file. (We needed this option as we used an embedded asset for browscap.ini via go-bindata.)
• Add test for initializing from Reader.
• Improve current tests, which rely on a particular order of execution in order to pass; namely, they assume that the library has been initialized first. This change ensures that the tests still pass when run individually.